SALINE AQUIFER CO2 STORAGE PROJECT (SACS)

BEST PRACTISE MANUAL

Report Number PH4/21 July 2003

This document has been prepared for the Executive Committee of the Programme. It is not a publication of the Operating Agent, International Energy Agency or its Secretariat.

SALINE AQUIFER CO2 STORAGE PROJECT (SACS) BEST PRACTISE MANUAL

Acknowledgement

The SACS Best Practise Manual was prepared by the SACS Project Team. The Best Practice Manual has been reproduced by the IEA Greenhouse Gas R&D Programme for dissemination amongst its members with the kind

permission of Statoil, the SACS project managers, and the SACS project partners.

i

SALINE AQUIFER CO2 STORAGE PROJECT (SACS) BEST PRACTISE MANUAL

Background In 1996, injection of CO2 into a deep saline aquifer, known as the Utsira Formation1, began in the North Sea. The CO2 is stripped from the produced natural gas on the Sleipner West field2. The CO2 injection rate since 1996 has been close to 1 Mt/y. The Sleipner project thus became the first commercial scale CO2 storage project in the world. In 1996, the Sleipner CO2 injection project offered a unique opportunity to study the fate of the CO2 injected into the Utsira formation. To take advantage of this opportunity the IEA Greenhouse Gas R&D Programme and Statoil organised a workshop in Trondheim, Norway in 1997. The workshop was successful in bringing together a group of industrial partners and research organisations who together developed a research programme to monitor and study the fate of the injected CO2

3. As a result of the workshop the Saline Aquifer CO2 Storage project, otherwise commonly known by the acronym SACS, was born. The SACS Project The SACS project has been supported by a number of industrial companies and the European Commission4. The main industrial funding partners are: Statoil, BP, ExxonMobil, Norsk Hydro, Vattenfall and TotalFinaElf. The research work has been undertaken by a number of research providers from Europe. The research providers included: British Geological Survey (BGS), Bureau de Geologiques et Minerers (BRGM), Geological Survey of Denmark and Greenland (GEUS), Institute Francais du Petrole (IFP), Netherlands Geological Survey (NITG-TNO), SINTEF Petroleum Research, Schlumberger Research and Nansen ERS Centre. The IEA Greenhouse Gas R&D Programme has also participated in the SACS project assisting in dissemination of project results and helping to facilitate international co-operation in the project. In addition, IEA Greenhouse Gas R&D Programme has assisted in organising collaborative R&D activities between the SACS project and a number of non European research groups including the GEODISC project in Australia and LLNL5 in the USA. The SACS project was completed in three phases, which were:

Phase 0 Baseline Data Gathering and Evaluation Phase 1 Establishment of Project Status after three years of CO2 Injection. Phase 2 Data Interpretation and Model Verification.

1 The Utsira Formation is a large salt water containing sand layer which lies 1000 meter below sea bottom. 2 A special feature of gas from Sleipner West is the high content of carbon dioxide (9% by volume) which must be reduced to 2.5% by vol. to meet pipeline standards to allow transportation of the natural gas onshore. 3 A report of the workshop entitled the Sleipner Carbon Dioxide Storage Workshop was prepared. The report number was Ph3/1, February 1998. 4 Phase 1 of the project was supported under the Thermie programme of the European Commissions 4th Framework programme. Phase 2 was supported under the European Commissions 5th Framework Programme. 5 Lawrence Livermore National Laboratory

ii

The first of Phase of SACS (Phase 0) was funded exclusively by the project partners. The Phase 0 work programme (1 Year) was completed in November 1998. The data gathering project was undertaken before the main SACS project began. The study produced a comprehensive report (for use by the project) on the baseline geologic data on the Utsira formation and on other equivalent formations in the North Sea. The Phase 1 programme was a 2-year study that ended in December 1999. Phase 2 (or SACS2) of the project ended in December 2003. The combined learning from the Phases 1 and 2 of the SACS project has been synthesised into a Best Practise Manual. The manual aims to describe what was done, what was learnt, what went well and where are the perceived gaps in knowledge or data in the SACS project. The objective of the manual is to assist the development of new CO2 storage projects through knowledge sharing and help them to build on the experiences and lessons learnt in the SACS project. SACS Best Practise Manual The SACS Best Practise Manual consists of two parts. The first part outlines the operational experiences gained during the Sleipner CO2 injection operation. The second part consists of recommendations based on the experiences gained from monitoring the Sleipner CO2 injection operation during the SACS project. Part 2 is divided into 4 chapters, corresponding to the 4 main work areas of the SACS project which are:

• Characterisation of the reservoir and cap rock, • Monitoring the CO2 injection process, • Reservoir simulation, • And geochemical characterisation.

The key conclusions and recommendations from the Best Practise Manual are summarised below. The detailed discussion leading to these conclusions and recommendations is contained in the main report. Characterisation of the reservoir and cap rock The project recommends that a detailed characterisation of the reservoir and cap rock, both on a local and regional scale, is essential before injection commences. The characterisation should involve a determination of structure and stratigraphy both within and external to the reservoir, together with the physical properties of both the reservoir and cap rock. Some of the key features of the characterisation programme include: • Identification and mapping of any faults in the reservoir and cap rock (including an

assessment of fault sealing capacity and potential for reservoir compartmentalisation) to determine the long term fate of the injected CO2.

• Determination of reservoir properties, such as porosity and permeability, to quantify potential storage capacity and likely migration paths and rates. To determine these properties, it is of the greatest importance to obtain core samples from the reservoir close to the injection point.

• An assessment of the total reservoir storage potential is desirable, so that a proper injection strategy can be devised.

• An assessment of the natural fluid flow in the reservoir is essential because of its potential to affect the migration of CO2.

• Also, it is considered that injection-induced pressure changes could lead to compromise of the cap rock seal, hence possible geo-mechanical consequences should be assessed prior to the start of injection.

iii

It is noted that experience from SACS indicates that the accuracy to which the structure needs be resolved depends on the structural form of the reservoir into which the CO2 is injected. If injection is into a large dome-like structure with a seal or closure of thickness several tens of metres or more, CO2 migration is likely to be well constrained and small uncertainties in reservoir geometry will be relatively insignificant. If, on the other hand, the injection is into a reservoir with gentle dips in the cap rock and only minor changes in topography at its top (as occurs in the Utsira Formation), much more detailed investigation of the depth profile of the reservoir will be required. This more detailed mapping is needed to permit the accurate definition of the structure of the top surface of the reservoir, to allow the prediction of the overall migration direction and evaluation of the location and volume of any structurally-defined traps along the migration paths. Monitoring the CO2 injection process One of the major successes of the SACS project was the demonstrated ability to monitor the injected CO2 using repeat-seismic surveying. Even with the CO2 in a supercritical, rather than a gaseous, state it has been shown that CO2 accumulations of thickness as little as one metre can be detected - far below the expected seismic resolution of approximately 7m. Even thin accumulations of CO2 were found to cause significant, observable and measurable changes in the seismic signal, both in amplitude and in travel time. Other monitoring techniques for offshore application have been assessed as part of the SACS project. These techniques included:

• Multi component (MC) seismic monitoring, • Gravity surveying, • Micro-seismic monitoring.

Whilst both MC seismic and gravity surveying have potential benefits, cost is considered to be a severe drawback for offshore monitoring. For example, offshore a typical MC data set may be between 5 and 10 times as expensive as a conventional 3D survey. Micro-seismicity was also considered to have potential as a continuous monitoring tool. However, before it could be considered as a monitoring too there was a need to determine first whether micro-seismicity actually does exist in the Utsira formation6. Reservoir simulation The SACS project would recommend that, for any geological CO2 storage projects, pre-injection reservoir simulation should be carried out with a reservoir model which is based on the best available geological data. Reservoir simulations can predict the CO2 injection rate that could be maintained, the rise in reservoir pressure caused by the injection, the likely lateral migration of the injected CO2 and the potential for CO2 dissolution into the formation water. The SACS work has shown that existing reservoir simulators (such as Simed II and Eclipse 100) can be applied to model CO2 migration provided that the physical properties of the CO2/brine system are well represented. For all time-scales the density difference between brine and CO2 and the viscosities are the dominating fluid parameters.

6 Micro-seismicity is more typical in low porosity carbonate rocks than in sand bodies like the Utsira formation.

iv

Also SACS has shown that by monitoring the CO2 distributions using repeat-3D-seismic surveys, quantitative information describing the formation can be provided by calibrating a simulation model to the development of the seismic images. This information can be used to build larger reservoir models, which can be used for long-term predictions. Geochemical characterisation The SACS project considers that it essential to have a good understanding of the fluid chemistry and mineralogical composition of any potential reservoir and cap rock so as to elucidate their reactivity with CO2. The degree of reactivity between CO2, pore water and minerals will influence the long-term storage potential of the formation. Depending on the nature and scale of the chemical reactions, the reservoir-CO2 interactions may have significant consequences for the CO2 storage capacity, the injection process, and long-term safety, stability and environmental aspects of CO2 storage.

The fluid chemistry and mineralogical composition of reservoir and cap rocks will be site specific, so it is considered to be important to recognise that geochemical investigations need to be carried out on a site-to-site basis. Based on the experience drawn from SACS, it is concluded that it is essential to determine the baseline geochemical conditions prior to CO2 injection and then determine the geochemical impact of injected CO2. In addition, it was concluded that analysis of borehole core material from the cap rock is the only way to provide sufficiently detailed information on cap rock mineralogy and pore water chemistry. The results from SACS indicated that the Utsira sand showed only limited reaction with CO2. Most reaction occurred with carbonate phases (shell fragments), but these were a very minor proportion (about 3% by weight) of the overall solid material. Discussion The development of a Best Practise Manual (BPM) by the SACS project will clearly contribute to the development of our knowledge and understanding of storing CO2 underground in geological formations. By developing such a document this will also assist in knowledge sharing that should benefit new CO2 storage projects. There is also considerable value in comparing “best practise” in other sectors such as acid gas injection, natural gas storage and from CO2/EOR operations in North America. A detailed geological characterisation of reservoirs prior to injection is already a feature of other operations. For instance it is embodied within the regulations for acid gas injection operations in Canada7 and for natural gas storage (NGS) operations in Europe8. Geochemical considerations are also included in assessments for acid gas injection operations because of the solubility of the acid gases (CO2 and H2S), but are not included for natural gas storage because the methane will not react or dissolve in the formation/formation water. Missing from the SACS BPM compared to the NGS activities, is the need to characterise overlying aquifers to determine the potential for potable water pollution if leakage occurs. That would be a more significant component of an

7 Acid Gas Injection: A Study of Existing Operations. Phase I: Final Report, Report Number Ph4/18, April 2003. 8 Geological Storage of CO2; Conflicts and experience to be gained from natural gas storage, File Note 4th January 2001.

v

onshore storage operation, whilst the SACS BPM limits itself to the experiences of offshore operations based around the Sleipner CO2 injection. The assessment of the cap rock sealing characteristics is another prerequisite in other storage operations such as NGS. Whilst, it is noted that a cap rock core was not obtained prior to CO2 injection at Sleipner, it has been subsequently taken and the need to take a core prior to injection operations has been emphasised in the BPM. In addition, it is recognised that abandoned exploration/production wells represent a potential early-release path for the injected CO2, if they have not been abandoned with due regard to subsequent injection of CO2. Any assessment of the reservoir integrity should include an assessment of any wells drilled through the storage structure. In well developed onshore regions like North America, the number of wells drilled is much greater than in the North Sea and is likely to present a greater risk. Detailed reservoir modelling is also a requirement in other storage operations; again it is specified in the European Gas Storage regulations, EN1918, as is an assessment of the geo mechanical properties of the cap rock. Such as assessment, should include the impact of earlier production operations if applicable, although in the case of Sleipner this was not applicable. One area where the stipulations of the SACS BPM differ from other storage operations is in the area of monitoring. In acid gas injection operations, little subsurface monitoring is undertaken, except for bottom hole pressure monitoring. For NGS operations monitoring wells are a specified feature of the regulations in Europe. In European gas storage operations, monitoring wells are normally sited in the reservoir summit and in an overlying aquifer to monitor for leakage and to monitor lateral gas migration in areas either up-dip of the storage reservoir or where permeability effects could result in preferential migration. It is noted that natural gas is a valuable commodity and so best efforts are normally made by the storage operators to ensure that loss does not occur. Of course, there are also safety implications because of the flammability of natural gas. The SACS project has identified repeat-seismic surveying as its technically preferred and cost effective monitoring option offshore. However, it is uncertain whether regulatory bodies in Europe, and the general public, would have sufficient confidence in repeat-seismic surveying alone, particularly for the first such storage operations. As confidence grows as a result of continued seismic monitoring at Sleipner, and as results are obtained from other monitoring projects (such as the Weyburn Monitoring Project), acceptance of seismic monitoring may develop in the future. It is possible that, in early commercial CO2 storage operations in Europe, a combination of observation and seismic monitoring may be required, although this will undoubtedly raise the cost of monitoring CO2 storage operations considerably. A distinction has been drawn here between Europe and North America – in the latter region, less regulatory emphasis is currently placed on subsurface monitoring for injection operations, although this may change in the future. An area not covered in the SACS BPM is risk assessment (it is a component of the follow-on project CO2STORE9 which will aim to apply the knowledge developed in SACS to 4 new reservoirs in North Western Europe). It is recognised that the understanding of best practise will develop further through CO2STORE. Risk assessment of storage reservoirs is an established technique and a number of research groups are now adapting existing methodologies to

9 A new European Commission supported project, called CO2STORE, commenced in February 2003 that will extend the work on the Utsira Formation to investigate the long term fate of the injected CO2 and evaluate other monitoring techniques, and to apply the knowledge gained in SACS to develop site-specific plans for CO2 storage operations elsewhere in Europe, both onshore and offshore

vi

geological storage of CO2. Assessing the risk of CO2 migration out of any storage reservoirs, if undertaken as part of a structured development plan for a storage project, could help boost public confidence in the technology as a safe global mitigation option. Conclusions and Recommendations The SACS Best Practise Manual is a useful reference document based on the experiences gained from the first commercial CO2 injection project in the North Sea and its accompanying monitoring programme. It is hoped that other projects now underway will document their findings in a similar way to share information globally and help build confidence in the CO2 storage technology. A limited review of best practise in related storage activities has been provided here. A more extensive review should be considered – this could cover not just acid-gas injection and natural gas storage operations but also operational experience with CO2/EOR and experience from water/gas injection operations in the North Sea. This would help to develop a body of knowledge useful for building confidence in CO2 storage in geological formations.

SALINE AQUIFER CO2 STORAGE PROJECT BEST PRACTISE MANUAL

1st Edition MARCH 2003

Statoil BGS BP BRGM ExxonMobil GEUS Norsk Hydro IFP Vattenfall NITG-TNO

SINTEF

IEA Greenhouse Gas R&D Programme Geco-Prakla

Nansen Research Centre

FOREWORD The SACS project is a research and demonstration project which is monitoring and forward modelling the underground CO2 sequestration operation taking place at the Sleipner West gas field, offshore Norway. The Sleipner project is the first and still the only CO2 sequestration project in the world being undertaken for environmental purposes. Thus CO2 sequestration is very much in its infancy, when Best Practice is mainly concerned with knowledge sharing and building on the experiences and lessons learnt in pioneering projects. With this in mind, the aim of this manual is to describe what we did in the SACS project, why we did it, what we learnt, what went well and where we perceive gaps in our knowledge or data. The manual is based entirely on our experiences of monitoring the CO2 sequestration operation at Sleipner. When considering its application to other potential sequestration sites, it is important to bear in mind that the Earth's subsurface is an extremely variable natural system and its properties are highly site specific. Thus, the importance of some of the issues and procedures highlighted in this manual will vary between sites and, as new potential sites and sequestration concepts are investigated, they may throw up issues not considered important at Sleipner. This means that the manual should not be regarded as a set of standard procedures for the investigation or monitoring of a potential CO2 sequestration operation. The manual consists of two parts. Part 1 is a short description of the Sleipner CO2 sequestration operation. Part 2 consists of recommendations based on our experiences of monitoring the Sleipner CO2 sequestration operation during the SACS project. Part 2 is divided into 4 chapters, corresponding to the 4 main work areas of the SACS project; characterisation of the reservoir and cap rock, monitoring the CO2 injection process, reservoir simulation and geochemical characterisation. Where possible, links to publications arising from the SACS project, and to other publications, have been included.

i

LIST OF CONTRIBUTORS The Sleipner CO2 injection operation Bjorn Berger Statoil Geoscientific characterisation of the reservoir and cap rock Andy Chadwick BGS Gary Kirby BGS Sam Holloway BGS Jonathan Pearce BGS Simon Kemp BGS Ulrik Gregersen GEUS Peter Johannessen GEUS Lars Kristensen GEUS Torben Bidstrup GEUS Rob Arts NITG-TNO Peter Zweigel Sintef Petroleum Research Anne-Elisabeth Lothe Sintef Petroleum Research Monitoring the injection process Rob Arts NITG-TNO Bert van der Meer NITG-TNO Ola Eiken Statoil Ivar Brevik Statoil Andy Chadwick BGS Gary Kirby BGS Sam Holloway BGS Peter Zweigel Sintef Petroleum Research Svend Ostmo Sintef Petroleum Research Emmanuel Cause Sintef Petroleum Research Anne-Elisabeth Lothe Sintef Petroleum Research Hubert Fabriol BRGM Bernard Zinszner IFP Trygve Randen GECO Kjetil Fagervik GECO Magne Lygren GECO Reservoir simulation in SACS: Verifying the seismic and geological interpretations and predicting the long-term fate of CO2 Erik Lindeberg Sintef Petroleum Research Per Bergmo Sintef Petroleum Research Bert van der Meer NITG-TNO Assessing the geochemical effects of CO2 injection Isabelle Czernichowski-Lauriol BRGM Bernard Sanjuan BRGM Christophe Kervevan BRGM Helene Serra BRGM Christopher A. Rochelle BGS Keith Bateman BGS Jonathan M. Pearce BGS Y.A. Moore BGS Niels Springer GEUS Christian Hoier GEUS Holger Lindgren GEUS Etienne Brosse IFP Sandrine Portier IFP

ii

FOREWORD.................................................................................................................... i PART 1: The Sleipner CO2 injection operation........................................................... 1

The Sleipner fields................................................................................................... 1 The CO2 removal process........................................................................................ 1 The CO2 injection operation.................................................................................... 2

CO2 tax……................................................................................................... 2 Disposal alternatives ...................................................................................... 3 Injection planning .......................................................................................... 3 The well ........................................................................................................ 3 Injected CO2................................................................................................... 4

PART 2: Recommendations based on monitoring the Sleipner CO2 sequestration operation during the SACS project ............................................. 5

Geoscientific characterisation of the reservoir and caprock ...................................... 5 Datasets available to the SACS project .................................................................. 5 Characterisation of reservoirs.................................................................................. 7

Reservoir Structure ........................................................................................ 7 Reservoir properties ....................................................................................... 8 Effective Storage Capacity........................................................................... 10 Natural fluid flow......................................................................................... 11

Characterisation of caprocks ................................................................................. 11 Conclusions and recommendations ....................................................................... 16

Monitoring the injection process ................................................................................. 17 Time lapse seismic data......................................................................................... 17 Assessment of multi-component (MC) seismic monitoring.................................. 22 Assessment of gravity surveying as a monitoring tool.......................................... 23 Assessment of microseismic monitoring............................................................... 23 Petroacoustics and thermodynamics related to seismic monitoring...................... 24 Integration of time-lapse seismic with reservoir flow model................................ 25 CO2 volume estimation from seismic data ............................................................ 26

Reservoir simulation in SACS: Verifying the seismic and geological interpretations and predicting the long-term fate of CO2 ............................... 27 Introduction ........................................................................................................... 27 Calibration of a local reservoir model by use of repeated 3D seismic.................. 27

Reservoir model ........................................................................................... 27 Fluid and transport properties ...................................................................... 28 Simulation tools ........................................................................................... 29 Results of model calibration ........................................................................ 30

Simulation of the long-term fate of CO2 in a large-scale model ........................... 33 Conclusions and recommendations ....................................................................... 38

Conclusions.................................................................................................. 38 Recommendations........................................................................................ 38

Assessing the geochemical effects of CO2 injection.................................................... 39 Determination of baseline geochemical conditions prior to CO2 injection........... 39

Reservoir formation ..................................................................................... 40

iii

Reservoir seal............................................................................................... 42 Determination of the geochemical impact of injected CO2................................... 42

Observations from laboratory experiments, field monitoring and natural analogues ........................................................................... 43

Numerical modelling.................................................................................... 45 Conclusions and recommendations ....................................................................... 46

References...................................................................................................................... 51

iv

PART 1: The Sleipner CO2 injection operation

The Sleipner fields

The Sleipner fields are located in the Norwegian sector of the North Sea, about 250km from the west coast of Norway (Figure 1).

Figure 1. Location of the Sleipner fields

The production licences that cover these fields are operated by Statoil. The licence partners are: ExxonMobil, Norsk Hydro and TotalFinaElf. The main reserves are gas/condensate in the Sleipner East and Sleipner West Fields.

According to official Norwegian government sources the Sleipner West Field originally contained 202 GSm3 of rich gas. The rich gas has a CO2 content that varies from 4 to 9.5 %. To be able to deliver Sleipner West natural gas directly into the gas pipelines to Europe, the CO2 content has to be reduced below 2,5%. To meet this specification CO2 is removed from the natural gas.

The Sleipner West Field is produced from a wellhead platform through a 12 km long pipeline to the process and treatment platform Sleipner T that is bridge connected to the Sleipner A platform on the Sleipner East field (see Figure 2). Production started in August 1996.

The CO2 removal process

The CO2 is removed from the natural gas by an active amine process in one of the processing modules at the Sleipner T platform. . A co-operation agreement has been concluded with Elf, which holds the patent for the system. Natural gas to be treated

1

passes through two large absorption columns, weighing about 8 000 tonnes and standing 35 m high. Energy freed by the amine process runs two generators, yielding 6 megawatts of power for use on the platform. After the carbon dioxide has been separated it is transferred to the Sleipner A platform for injection. The natural gas is treated in another module and exported via Sleipner A through the pipeline system to continental Europe. Some of the natural gas also is injected into the Sleipner East reservoir to improve condensate production. The cost of the carbon dioxide module was about NOK 2 billion.

The CO2 injection operation

CO2 tax As early as in January 1991 a tax on CO2 emissions from fossil fuel combustion offshore was introduced in Norway. Presently this tax is NOK 0.72 per Sm3 of gas, which corresponds to NOK 400 per metric tonne of CO2. During the planning of the Sleipner West field development, levels of environmental concern increased significantly. Releasing approximately 1 million metric tons per year of CO2 to the atmosphere would have represented a 3% increase in the total Norwegian CO2 emissions. So it was felt that even if it was not coming from fossil fuel combustion, the CO2 from the gas processing operation at Sleipner might become subject to CO2 tax.

Figure 2. The Sleipner T (right foreground) and Sleipner A platforms

In an overall effort to reduce the emission of combustion gases to air, the Sleipner West licence partners therefore decided not to vent the CO2 to the atmosphere.

2

Disposal alternatives As thoroughly described by (Baklid, Korböl and Owren, 1996), the following alternatives for disposal were investigated:

• Export by pipeline to an oilfield in the area for enhanced oil recovery purposes • Injection into the main Sleipner East reservoir for enhanced condensate recovery • Injection into the Heimdal Formation for disposal • Injection into the Utsira Formation for disposal

For the first alternative the amounts of CO2 needed in the possible fields did not match the production rate. For the two next the risk of unwanted CO2 contamination of the produced gas from Sleipner East was considered too high. So the Utsira Formation was chosen.

Injection planning Based on a structural contour map of the Utsira Formation in the Sleipner East area a reservoir simulation study was performed with a “modified” black oil simulator using the following physical data: Pressure 8 – 10 MPa Temperature 370 C Permeability 1- 8 Darcy Porosity 35-40% Net sand 80-100% The results of this study indicated that:

• The CO2 should be injected near the bottom of the formation to minimise areal distribution and maximise the dissolution in formation water

• The maximum extension of CO2 after 20 years of injection would be 3 km • There were no major differences in areal distribution between free and dissolved

CO2. • Supercritical conditions resulted in a wider distribution of free CO2. • Up to 18% of the CO2 injected was dissolved in the formation water

The CO2 is injected below a small structural closure northeast of the SLA platform. The CO2 gas - not dissolved in formation water - is expected to be captured in this closure. When the closure is filled, the CO2 is expected to spill towards the north and thereafter to the northwest. The CO2 gas is not expected to reach the SLA platform location and thus corrosion on wells due to the injected CO2 is avoided

The well For details about design of the well refer to Baklid, Korböl and Owren (1996). To assure the necessary lifetime of 25 years, 25% Cr duplex (stainless) steel was chosen for the well. The design capacity was set to 1.7 MSm3 of CO2 per day and to keep the CO2 in the dense phase. In the injection area the 7” casing was perforated over a length of 100m to

3

achieve this capacity. In the early phase of the injection there were some problems due to collapse in the unconsolidated sand. Installing a sand screen restored the well injectivity.

Injected CO2 The cumulative amount of CO2 injected, and the wellhead pressure is shown in Figures 3 and 4 below:

Injected CO2

0

500000000

1000000000

1500000000

2000000000

2500000000

3000000000

13-Sep-96 13-Sep-97 13-Sep-98 13-Sep-99 12-Sep-00 12-Sep-01

Date

Wellhead pressure

0102030405060708090

13-Sep-96 15-Mar-97 14-Sep-97 16-Mar-98 15-Sep-98 17-Mar-99 16-Sep-99 17-Mar-00 16-Sep-00 18-Mar-01 17-Sep-01

Date

Wel

lhea

d pr

essu

re (b

ar)

Figures 3 and 4. Cumulative injected CO2 and well head pressure during the Sleipner CO2 injection operation to February 2002.

4

PART 2: Recommendations based on monitoring the Sleipner CO2 sequestration operation during the SACS project

Geoscientific characterisation of the reservoir and cap rock It is necessary to characterise the reservoir and cap rock on both local and regional scales to elucidate CO2 migration patterns and overall storage potential. This involves a determination of structure and stratigraphy both within and external to the reservoir, together with the physical properties of both the reservoir and cap rock. In the following description we aim to draw out general recommendations for the different aspects of characterisation, based on our experience during the SACS project, where geoscientific appraisal of the reservoir (the Utsira Sand) and its cap rock was carried out at a range of scales. The whole reservoir (some 26000 km2) was mapped and characterised using regional 2D seismic datasets and well data. More detailed work was carried out around the injection site using a 3D seismic dataset and more closely spaced well data. Key issues discussed are listed below: • Datasets • Characterisation of reservoirs

Reservoir Structure Reservoir Properties Effective Storage Capacity Natural Fluid Flow

• Characterisation of cap rocks

Datasets available to the SACS project The following datasets were available for the SACS regional study: • Seismic data • 2D seismic profiles ~ 16000 line km

3D seismic volume ~ 770 km2

• Well data (~300 wells, ~ 30 within 20 km of the injection site) Lists of formation tops Geophysical logs Reservoir core material Selected cuttings of cap rock and reservoir 3 reservoir pressure measurements (2 from Sleipner, 1 from Brage, ~ 250 km to

the north)

5

Figure 5 a) Typical 2D seismic reflection profile across the Utsira reservoir b) Regional depth map to top Utsira Sand based on 2D seismic surveys and incorporating 3D data around Sleipner injection point. c) Detailed depth map of Top Utsira Sand around Sleipner injection point (IP), based on 3D seismic data. [2D seismic data courtesy of Schlumberger Geco-Prakla]. The 2D and 3D seismic data constituted the key datasets, essential for delineating the reservoir limits, structure and stratigraphical correlation (Figure 5a). The regional seismic

6

datasets were proprietary, so SACS had no control over acquisition and processing parameters. 2D data quality was variable, ranging from moderate early 1980's data to good/very good late 1980's data. The 3D data were of good quality. In fact regional reservoir mapping is relatively insensitive to data quality, and, in the general case, cheaper, older datasets may offer good value for money. Careful assessment of data and requirements is recommended prior to purchase or acquisition. From the point of view of SACS, the large number of wells was useful for delineating regional structure, and was essential for mapping reservoir properties, such as porosity. In general terms however, it is considered that a lesser amount of well data, perhaps only 20% of the available dataset, would have been adequate for the purposes of the SACS regional investigations. Concentration near to the injection point is of course ideal. Of the available geophysical logs, the γ-ray log was the most useful general-purpose tool for identifying the reservoir sand and quantifying sand/shale ratios, augmented by the resistivity log. Sonic and density logs were utilised for porosity determination and mapping. In general terms, a reduction in the amount of the seismic data, depending on location, could have significantly reduced the confidence of the regional reservoir mapping. A proportionately similar reduction in well data would not have had such a seriously detrimental effect on the regional mapping, but would have adversely affected confidence in reservoir characterisation and storage estimates, particularly around the injection point.

Characterisation of reservoirs It is considered that the properties of a reservoir in the subsurface can best be determined by an analysis of seismic and borehole data augmented by rock material (core and cuttings). This should aim to produce information on structure, stratigraphy and physical properties. The mapping should include, as a minimum, depth to top reservoir, reservoir thickness and reservoir physical properties (porosity and sand/shale ratio if appropriate). It is also essential to understand the lateral and vertical stratigraphical and hydraulic continuity of the reservoir.

Reservoir Structure It is recommended that the following datasets are available for determination of the structure of the reservoir: • A regular grid of 2D seismic data over the entire reservoir. • A high quality 3D seismic volume over the injection site and adjacent area, tuned if

possible for optimal resolution at the storage depths. • Sufficient borehole data to permit accurate and constrained depth conversion of the

seismic data, particularly in the case of a flat or near flat-lying reservoir. As CO2 is buoyant (in both gaseous and fluid phases) it will tend to rise to the top of the repository reservoir. Assessment of the depth to the top of the reservoir is therefore a basic prerequisite of CO2 storage (Figure 5b). It allows a first order estimate of short-term storage capacity, and permits likely migration pathways and extents to be assessed. The

7

accuracy, to which structure needs be resolved, however, depends on the structural form of the reservoir into which the CO2 is injected. If injection is into a large domal structure with closure of several tens of metres or more, CO2 migration trends are likely to be well constrained and small uncertainties in reservoir geometry are not significant. If, on the other hand, the injection is into a reservoir with gentle dips and only minor topography at its top (as at Sleipner), very detailed depth mapping is required (Figure 5c). This will permit accurate definition of the structure of the top surface to allow the prediction of the overall migration direction and evaluation of the location and volume of any structurally defined traps along the migration paths. Detailed subsurface structural mapping requires a minimum of a regular grid of 2D seismic and preferably a 3D seismic survey around the injection site. In the case of very low structural relief it is essential to produce an accurate depth map. This requires sufficient velocity control from nearby boreholes to effectively minimise uncertainties in depth conversion. The Sleipner case provides a good example. The top of the reservoir above the injection point is gently undulating but relatively flat. Uncertainty of just a few metres in regional depth trends (requiring less than 1% error in depth conversion) will radically alter the modelled migration direction (see below). This has impacts not only on assessing the long-term safety-case, but also, from a practical point of view, the design of future monitoring surveys. Although significant faulting has not been identified so far in the Sleipner CO2 repository, in the general case it is important to identify and map any faults in the reservoir and cap rock, and to make some assessment of fault sealing capacity (e.g. by empirical fault gouge shale ratio estimation), so as to be able to detect and assess possible reservoir compartmentalization and/or the potential for fault-related leakage. The presence of reservoir compartmentalisation could lead to a rapid increase in formation pressures with time as fluid flow between compartments is inhibited. The identification of small-scale faulting (throws of a few metres) requires seismic data of adequate resolution. Fair quality 2D seismic should be sufficient to identify the presence of significant faulting, but accurate mapping of fault networks and linkages requires 3D data coverage.

Reservoir properties Knowledge of reservoir properties, such as porosity and permeability, is required to quantify potential storage capacity and likely migration paths and rates. To determine these properties, the importance of having core material from the reservoir close to the injection point cannot be overemphasized (Figure 6a, b). Core and cuttings material from additional wells will further improve characterisation, particularly if vertical and lateral reservoir in homogeneity is suspected. Determinations from material in the likely CO2 migration pathway, i.e. the top of the reservoir, are of particular importance. Analysis of the reservoir core should be prioritised according to the requirements of the reservoir (transport and reaction-transport) modellers, but is likely to include: • Sedimentology, petrography, fabric

Optical microscopy (optical porosity) SEM (scanning electron microscopy)

• Mineralogy

8

XRD (x-ray diffraction) Particle-size analysis

• Transport and sealing properties Absolute permeability, effective and relative permeability Porosity

• Reservoir-water-CO2 chemical reactions Pore water analysis Chemical reactions (dissolution/precipitation) Physical reactions (dehydration)

Figure 6 a) b) SEM images of Utsira Sand from core sample c) Well correlation diagram from the southern part of the Utsira Sand utilizing γ-ray and sonic logs (total section length about 85 km). Note how γ-ray logs resolve thin intra-reservoir shales (arrowed), and laterally variable sand/shale ratio. In order to extrapolate effectively from the coring point(s) it is necessary to have geophysical log data, suitable for physical property determination, from wells at least as far from the injection point as the predicted CO2 migration (Figure 6c). Wherever

9

possible, outcrop information should be incorporated into the characterisation process. Outcrop correlatives or analogues are valuable in understanding the nature of spatial variation in reservoir properties, and geostatistical or stochastic methods of 3D reservoir model building may be useful. The amount of information needed to characterise the reservoir varies with reservoir type. The Utsira reservoir forms a fine- to medium-grained poorly-consolidated sand, interpreted by SACS as a lowstand turbidite fan complex with stacked mounded sand bodies interspersed with thin laterally impersistent shales. In regional terms the fairly sparse cover of wells appears sufficient to characterise the reservoir adequately in terms of broad stratigraphy and storage capacity (Table 1A). This would not necessarily be the case with more complex reservoirs.

% mineral grain size porosity permeability sand/shale

ratio quartz calcite K-

feldspar albite aragonite mica and others

fine (medium)

35 - 40 % (27 - 42 %) 1 - 3 Darcy 0.7 - 1.0

(0.5 - 1.0) 75 3 13 3 3 3

Table 1A. Generalised properties of the Utsira Sand from core and cuttings. Mineral percentages based on whole-rock XRD analysis. Even here, though, considerable variation of porosity and sand-shale ratio are evident, of considerable significance when calculating the regional reservoir storage capacity. For SACS physical properties were mapped in 2D (x, y, value) across the entire reservoir unit. In other reservoirs, physical properties may vary more significantly and information from more wells is required to define the variability and to assess whether it is systematic or random. In some cases a full 3D reservoir property model (x,y,z,value) may be deemed necessary. To this end, an understanding of the environment of deposition of the reservoir is important as this will provide models for the likely distribution of different lithologies and therefore lateral variations away from any borehole provings. Depositional assessment relies both on the interpretation of borehole data (cores, cuttings, logs) and on seismic (sequence) stratigraphic analysis. The latter may, in addition, provide specific details on the presence and geometry of internal migration barriers (e.g. shaly units on clinoforms in deltaic successions). The effect of internal flow barriers (either dipping or horizontal) on CO2 migration could be substantial in altering the migration path from the injection point to the top of the reservoir.

Effective Storage Capacity Assessment of the total reservoir storage potential is desirable, so that a proper injection strategy can be devised. Since CO2 is likely to pond buoyantly at or close to the reservoir top, it is important to distinguish between the component of reservoir storage that is enclosed within structural and/or stratigraphic traps (where CO2 can be expected to accumulate long-term) and the total theoretical storage volume of the reservoir (commonly a much higher value), which cannot necessarily be effectively utilised. The relationship between these two quantities is complex and varies from case to case. It

10

depends both on the structural geometry of the reservoir and also on its internal complexities (such as shale layers or other permeability barriers). It is important to determine the internal stratigraphy of the reservoir on a site specific basis. For example, in the Sleipner case the presence of thin shale beds is radically affecting CO2 distribution in the reservoir, with CO2 migrating laterally for several hundred metres beneath intra-reservoir shales (see below). It is likely that in the longer term this dissemination of CO2 throughout the reservoir thickness (rather than just being concentrated at the top) may allow more efficient dissolution of CO2 and effectively increase the reservoir capacity well above the minimum value defined by the volume of the top reservoir traps. None of these thin shale beds were clearly resolved on the seismic data (not even on the 3D data) and require geophysical well logs for their identification (even utilising log data, the thinner shales are below the thickness resolution limit). A suitably close coverage of wells is needed to allow well-to-well correlation and to obtain some understanding of shale continuity; in the Utsira Sand, firm correlation of shale beds was not possible even between wells a few (<10) kilometres apart. Biostratigraphically the Utsira Sand is poorly-understood. In other cases however, particularly carbonate reservoirs with more core material, high resolution biostratigraphy may have an important role in defining detailed reservoir models.

Natural fluid flow Natural fluid flow in the reservoir is a factor with the potential to affect the migration of CO2. Because CO2 has very low viscosity, even very small hydraulic pressure gradients can cause significant displacement of the CO2 plume, particularly if the reservoir topography is gentle. A significant natural fluid flow could cause migration paths to deviate significantly from those indicated by just a consideration of buoyancy effects. Fluid flow may be determined from physical measurements of pressure at different locations in the reservoir, or if such data is not available, by basin modelling software. In the case of the SACS project, available data indicate that the lateral pressure gradient in the Utsira Sand is very small, perhaps compatible with natural fluid flows in the order of 0.3 to 1 myr-1. The pressure data are very sparse however (see above) and this figure must be treated with considerable caution. SACS also used basin modelling techniques to calculate theoretical flow velocities based on the compaction history of the Utsira Sand. Velocities of 2 – 4 myr-1 were obtained for the reservoir around Sleipner, though rather conservative (high) permeability’s were assumed. On the other hand, reservoir simulations suggest that hydrodynamic displacement of the CO2 plume is insignificant, indicative of very low rates of natural fluid flow.

Characterisation of cap rocks Knowledge of the extent, nature and sealing capacity of the cap rock is perhaps the key purely geological element in assessing and establishing the long-term safety case for the CO2 repository. Determination of the extent of the cap rock will rely on a regional spread of boreholes and on the grids of 2D and 3D seismic data. Sample material should be available (probably in the form of core) in sufficient quantity to undertake a detailed suite of analytical tests. The core material should ideally be in a location above the likely CO2

11

migration pathway or from a demonstrably analogous position. Analysis of the cap rock core should be prioritised according to the requirements of the geomechanical and reservoir (transport and reaction-transport) modellers, but is likely to include: • Sedimentology, petrography, fabric

SEM (scanning electron microscopy) X-ray screening N2BET

• Mineralogy XRD (x-ray diffraction) Particle-size analysis CEC (cation exchange capacity) TOC (total organic carbon)

• Mechanical properties Mohr-Coulomb behaviour Young’s modulus Drained bulk modulus Cam-Clay parameters Time-dependent creep Poisson’s ratio Acoustic velocity Permeability

• Transport and sealing properties Capillary entry pressure Absolute permeability, effective and relative permeability

• Caprock-water-CO2 chemical reactions Pore water analysis Chemical reactions Physical reactions (dehydration)

In the absence of core material, drill cuttings (preferably augmented by sidewall core material) are suitable for a limited range of analytical techniques such as petrography, SEM, XRD (Figure 7).

% mineral sand (>63 µm)

silt (2 - 63 µm)

clay (<2 µm) quartz k-spar alb calc mica kaol smect chlor pyr gyp hal sylv bar

CEC meq/100g

TOC (%)

0 - 5% 40 - 60% 45 - 55% 30 5 2 3 30 14 3 1 1 1 2 1 5 6.0 - 20.2 0.68 -

1.28 Table 1B Generalised properties of Utsira cap rocks, based on analysis of cuttings. Results from cuttings analysis (e.g. Table 1B) can be used to assess sealing capacity in a qualitative manner, by comparison with samples from proven oil/gas field cap rocks, or semi-quantitatively such as by the Krushin grain-size method (Krushin, 1997).

12

Figure 7. Analysis of cuttings material from the Utsira cap rock succession. a) SEM image of massive mud rock with a number of rounded fine-grained quartz grains (arrowed). b) High magnification detail of laminated mud rock showing tightly packed platelets with preferred orientation. Micropores (arrowed) are a few microns in diameter and appear to be poorly connected. c) X-ray diffraction traces from two samples.

13

Figure 8. Migration pathways (purple) from the Sleipner injection point. a) Final distribution of 3 x 107 m3 (~ 20 MT) of CO2 assuming migration beneath the top of the Utsira Sand b) Final distribution of 7.4 x 106 (~5 MT) of CO2 assuming migration beneath the top of the sand-wedge. Note if more than 5 MT of CO2 it will migrate out of the area of 3D seismic coverage. Two-way time shading ranges from blue (deeper) to red (shallower).

In addition to establishing physical properties at a number of point locations (wells) it is necessary to evaluate the bulk geometry of the cap rock and any structures which may affect it (particularly in the vicinity of the predicted CO2 migration paths). The regional

14

seismic stratigraphy of the cap rock should be discernible from 2D seismic data, as would major faults that cut it. Smaller structural features for example ‘polygonal’ type minor faults that characterise some shale sequences, generally require 3D seismic data for their proper identification. Very small structures, fractures and joints are beneath the limit of seismic resolution. At Sleipner the cap rock succession is some 700 metres thick and is stratigraphically complex, comprising three main units (Figure 5a). The uppermost unit of Quaternary silts and muds overlies a thick dominantly silty Pliocene succession of prograding clinoforms. The lowermost unit comprises dominantly silty mudstone and seems to be basin-restricted. The ability of the seismic and well data to resolve fine stratigraphical detail around the reservoir/cap rock interface has proved essential to predicting potential migration patterns. It is likely that a thin sandy unit (termed the ‘sand-wedge’ by SACS) in the lowermost part of the cap rock will provide an important migration conduit; a small dip divergence between this and the top Utsira Sand results in an azimuthal change of some 90o in predicted migration direction (Figure 8). This has important consequences for migration modelling. In the Sleipner case, there is sufficient structural closure at the top of the Utsira Sand to trap 20 MT of CO2 within 12 km of the injection site (Figure 8a). However, if most of the CO2 migrates beneath the top of the sand-wedge the situation is less well constrained; only 5 MT of CO2 are sufficient for the migration stream to leaving the area of the 3D survey to the east (Figure 8b). This emphasises the need for very precise depth conversion when dealing with flat-lying repository aquifers. The inability of a cap rock succession to provide a long-term seal for the underlying reservoir may be revealed by indicators of hydrocarbon migration into and through the cap rock. Seismic amplitude anomalies and gas shows in the cap rock may signify the presence of shallow gas. Pockmarks and vents at the seafloor are indicators for gas migration from the underground into the sea water. However, gas within the cap rock may have formed biogenically in situ, and does not necessarily imply migration from below. The nature and source of shallow gas needs therefore to be addressed if indicators for its presence have been detected. The degree of correlation between seismically-imaged gas migration indicators and mapped faults is clearly of potential importance in evaluating fault-related leakage. At Sleipner however there is no clear spatial correlation between seismic amplitude anomalies and structure in the Utsira Sand. Injection-induced pressure changes could lead to compromise of the cap rock seal and possible geomechanical consequences should be assessed prior to injection commencing. Two main effects should be considered: fracture dilation due to increased pore-pressures and induced seismic slip due either to raised pore pressures or a reduction in normal stress due to buoyancy forces exerted by the CO2 plume. Fracture orientations that are likely to be conducive to fluid flow or susceptible to seismic slip can be determined relative to the principal stress axes if the in situ stress is known. At Sleipner, the required injection pressures (Figure 4) are considered most unlikely to induce either dilation of incipient fractures or microseismicity.

15

Conclusions and recommendations A recurring theme is that correct appraisal of structural and stratigraphical detail, at a range of scales, is very important. Small uncertainties in the geological model, particularly with regard to reservoir geometry and the presence of faulting, can lead to a striking variation in the predicted behaviour of CO2 in the reservoir, both in terms of migration paths and storage capacity. The Utsira Sand is a relatively simple and, in bulk terms, a laterally quite uniform reservoir. In other, more complicated cases, the construction of constrained, but simplified geological models based on all the available data may be required prior to reservoir simulation. This will necessarily require a substantial amount of time-intensive interaction between geoscientists and reservoir modelers, more so than was required for the SACS project.

16

Monitoring the injection process

Time lapse seismic data The major success of the SACS project has been the demonstration that conventional, time-lapse, p-wave seismic data can be a successful monitoring tool for CO2 injected into a saline aquifer (Eiken et al. 2000, Brevik et al. 2000, Arts et al. 2000). Even with the CO2 in a supercritical, rather than a gaseous, state it has been shown that CO2 accumulations with a thickness as low as about a metre can be detected - far below the conventional seismic resolution limit of approximately 7 m. Even these thin accumulations cause significant, observable and measurable changes in the seismic signal, both in amplitude and in travel time (Figure 9a).

Thin shale layer (2 m)

Base CO2 (0-8 m)

Seismic impedance model (in depth) withCO2 accumulation below a thin shale layer

Corresponding synthetic seismic traces (in time)

100

m100 m

s

MINIMUM AMPLITUDE of the tuned reflection

Pushdown below the CO2

Top Utsira Sand

Base Utsira Sand

Figure 9a: Synthetic model of a 2 m thin shale layer with an increasing CO2 accumulation (0-8 m) below. It is exactly this major effect on the time lapse seismic signal of relatively thin CO2 accumulations that has built our confidence that any major leakage into the overlying cap rock succession would have been detected. So far, no changes in the overburden have been observed in the Sleipner case. The time lapse seismic data have provided insights into the geometrical distribution of the injected CO2 at different time steps and show the different migration pathways. As expected, due to the lower density of CO2 with respect to the formation water, gravitational segregation is the dominant physical process governing the migration. The seismic data have revealed at least temporary barriers (very thin shale layers) to vertical migration of the CO2 that could not be resolved on the pre-injection baseline data alone. Due to the pronounced effect of the CO2 on the amplitude of the time lapse seismic signal these barriers have been mapped locally, markedly increasing our understanding of the CO2 migration within the reservoir. At various locations chimneys have been observed where CO2 passes through the thin shale layers (Figure 9b).

17

1994 1999 2001

1999-1994 2001-1994

Top sand wedge

Top Utsira Sand

CO2 in 1999 (yellow) CO2 in 2001 (green)

“Push-down” effect below the CO2 accumulation

Figure 9b: Inline 3838 through the injection area for the 1994, 1999 and the 2001 surveys including the difference between the 1999-1994 data and the 2001-1994 data. The interpreted CO2 levels are visualised in yellow (1999) green (2001). Interpretation of the post-stack seismic data has provided much of the information required to characterise the “CO2 bubble” including mapping the different CO2 levels and quantifying the amount of CO2 at each level (Fig. 10).

18

Top sand wedge Top Utsira Sand Level 1

Level 2 Level 3 Level 4

Level 5 Level 6 Level 7

Injection point

Inline 3838

CO2 distribution in October 1999Top sand wedge Top Utsira Sand Level 1

Level 2 Level 3 Level 4

Level 5 Level 6 Level 7

Injection point

Inline 3838

CO2 distribution in October 2001

2 km

Figure 10: Interpreted CO2 accumulations at different depth levels (amplitude maps from shallowest to deepest level 7). AVO analysis (or similar methods such as AVA or AVP) on the pre-stack data could be a tool to distinguish between saturation and pressure effects induced by the injected CO2. In the SACS case however no significant pressure build-up has been observed. Quantitative interpretation of the time lapse seismic data is necessarily linked both to the choice of an appropriate rock physics model, i.e. Gassmann (1951) and also to assumptions on saturation ranges and temperatures. By making these assumptions, a mass balance can be attempted by comparing the actual injected quantity of CO2 with the seismically derived quantity. Such an analysis has the potential to confirm (as a first order approximation) whether all of the CO2 is imaged by the time lapse seismic data. A reasonable match between the reservoir simulation model and the seismic data is required to gain insight in the predictive power of the reservoir simulation. With respect to the choice of the geometry of the time-lapse seismic surveys, these were a compromise between keeping the same geometry as the base survey and focussing more on the shallower target. In general this worked out satisfactorily. Some key elements of the acquisition and processing of the time-lapse seismic surveys for SACS are pointed out in Tables 1a and 1b.

19

Table 1a: Some key acquisition parameters for the SACS-seismic surveying

Survey ST9407 ST9906 ST0106 Data acquired 6.8-10.9 1994 8.10-10.10 1999 27.9-1.10 2001 Shooting direction 0.853 degrees 0.853 degrees 0.850 degrees Source tow depth 6 m 6 m 6 m No. of sub arrays 3 3 3 Source Xline separation

50 m 50 m 50 m

Source volume 3400 in3 3542 in3 3397 in3 No. of sources 2 2 2 Shot point interval 18.75 m 12.5 m 12.5 m No. of cables 5 4 4 (6) Cable separation 100 m 100 m 100 m Cable length 3000 m 3600 m 1500 m (3000 m) Distance from source to first group

195 m 165 m 150 m

Group interval 12.5 m 12.5 12.5 m Tow depth 8 m 8 m 8 m

20

Table 1b: Some key processing parameters for the SACS-seismic surveys

Survey ST9407 ST9906 ST0106 Select common offset range to process (approx. 240-1740 m)

yes yes yes

Deterministic zero phasing using supplied far-field signature

94-signature 99-signature 01-signature

Tidal correction based on model

yes yes yes

Swell noise attenuation using "Deband", applied in two frequency bands separated at 20 Hz

yes yes yes

Predictive deconvolution in tau-p space

yes yes yes

NMO, k-filter, spatial decim from12.5m to 25m receiver interval

yes yes yes

sort to 20 offset groups yes Resampling from 50 m to 75 m trace spacing

Resampling from 50 m to 75 m trace spacing

Swath dependent static correction based on offset group 4 & 5 (500-600m)

No Yes yes

Global amplitude scaling Factor 0.8264

Global frequency amplitude match

No Yes yes

Phase rotation -20 degrees Time shift -1.3 ms DMO Yes Yes Yes New NMO, with velocities slightly adjusted from a previous reprocessing of ST9407

yes New velocity picking in CO2 injection area

Same velocities as 1999

3D stack yes yes yes FX interpolation to 12.5 m x 12.5 m grid

yes yes yes

Finite difference migration with 100% interval velocities

yes yes yes

Residual frequency, phase and time match to ST9407

yes yes

21

Assessment of multi-component (MC) seismic monitoring Multi-component (MC) sensors can be used to record shear (S) waves as well as compressional (P) waves. On land, three polarised shear-wave sources, together with 3-component geophones, can be used to produce 9-component (full-wave) data. Offshore, 4C sea-bottom instruments (3 component geophones plus a hydrophone), utilise P to S mode conversions to record PS (P-downgoing, S-upgoing) datasets. MC datasets contain inherently more information than conventional data. Firstly, S-waves propagate exclusively through the rock matrix and are relatively unaffected (other than by pressure) by the nature of the pore fluid. This allows S-waves to image through volumes containing anomalous fluids (e.g. CO2 bubbles), more effectively than P-waves, and makes S-wave acoustic properties more uniquely diagnostic of lithology. Secondly, S-waves interrogate azimuthal subsurface properties, so the polarized waveform may exhibit birefringence due to velocity anisotropy. This can be used to measure azimuthal anisotropy in rock properties (due to structural fabric, or fractures) or lateral variations in effective stress (fluid pressure). To summarise, MC data can produce benefits in the following areas (Liu et al. (2001)): • Lithology, fluids and anisotropy • Lithological discrimination (sand/shale) • Fluid discrimination and saturation mapping • Fracture anisotropy and stress state characterisation • Improved imaging • Structural imaging through gas/CO2 cloud • Imaging beneath high velocity layers • Physical properties • Vp/Vs ratio (e.g. to help constrain Gassman equation) (properties in bold would be applicable to the Sleipner/Utsira injection operation). MC benefits are maximised in a fractured (or otherwise anisotropic) reservoir, particularly when the CO2 injection is accompanied by a significant increase (e.g. 5 – 10 MPa) in formation pressure. Neither of these conditions is satisfied at Sleipner, so full benefits cannot be realised. Cost is a drawback, most severely with offshore data (which is also technically compromised as discussed above). Land 3D MC data is typically 1.3 (3C) to 2.3 (9C) times as expensive as 3D P-wave data. Offshore however, a typical 3D PS dataset may be between 5 and 10 times as expensive as a conventional 3D survey.

22

Assessment of gravity surveying as a monitoring tool Monitoring the injected CO2 by repeated high-precision gravity measurements (micro-gravity) can provide better constraints on in situ CO2 density (Williamson et al., 2001). As seismic waves are fairly insensitive to density, gravity data can provide information which is complementary to that given by seismic methods. Such monitoring might be of particular use in mass and volume calculations. Thus if large quantities of CO2 dissolve in the formation water, this may be detected by gravity. Alternatively, if significant amounts of gas are breaching through the cap rock, gravity monitoring may serve as a “catastrophic early warning system”. The lateral resolution is much lower than for seismic monitoring, but for quantification and further dynamic modelling, it could, together with the seismic geometrical information, be a valuable additional monitoring tool, provided density contrasts are large enough. [In the Sleipner case with CO2 injected into the Utsira Sand, a detectable gravity change is expected to arise if CO2 densities are low (high geothermal gradient scenario)]. If such a monitoring project is undertaken, pre-injection baseline data are of great value and their acquisition is strongly recommended. At Sleipner, baseline gravity data were not collected, but the project is still considering the method for future monitoring. Offshore the only way to obtain sufficient accuracy is by seafloor measurements. Onshore, data acquisition is a lot less expensive, as no vessel is required. It is with onshore CO2 sequestration that gravity monitoring has the greatest potential benefit. Because of its cheapness, land gravimetry could be used to ‘interpolate’ between more widely-spaced (in time) seismic surveys to provide a cost-effective monitoring strategy.

Assessment of microseismic monitoring In general, the principal advantage of using microseismic monitoring is its continuous nature. In other words, if a cause and effect link can be established between the appearance of microseismicity and the increase in pore pressure in the reservoir due to the flow of CO2, then, theoretically, a real-time picture is provided of the passage of CO2 at certain specific points. It is also possible to characterize zones of weakness in the reservoir (or its cap rock), where pre-existing fractures or joints move in brittle shear and therefore constitute preferential flow paths. From a practical point of view, microseismicity appears mainly in low-porosity carbonate rocks and when injection pressures are relatively high (several tens of MPa). Given the porosity values at Sleipner, microseismicity is unlikely to appear in the Utsira Sand except perhaps in shale lenses or in the overlying shale cap rock (Fabriol, 2001). This latter case could be the most interesting to monitor as it would reveal the presence of leakage in the cap rock. However, it remains to be proven that microseismicity actually does exist in the Sleipner case. Clearly, under these conditions, microseismic monitoring is not the preferred tool for monitoring CO2 injection. However, in terms of drilling an observation well, we recommend:

23

• Install permanent geophones on the well casing for repeated VSP • Incorporate a system for continuously recording seismic background noise using

these sensors. With this recording setup, it would be possible to detect significant microearthquakes that are located within about half a kilometre of the well. This will allow the events to be correlated with either the intra-reservoir shales of the Utsira or those of the overlying cap rock. Though microseismic monitoring is not considered to be of great use at Sleipner, it is expected to be more appropriate to other CO2 underground storage projects, particularly in low permeability reservoirs.

Petroacoustics and thermodynamics related to seismic monitoring The quantitative interpretation of time-lapse seismic monitoring relies on a sufficiently accurate estimation of the fluid substitution impact on seismic velocity in the reservoir. The theoretical basis of this quantification is the well-known Gassmann (1951) model which relates the elastic moduli of the fluid saturated rock (bulk modulus: Ksat and shear modulus: µsat) • to the moduli of the dry rock (bulk: Kdry and Shear µdry), • to the saturating fluids bulk modulus (Kfl), • to the porosity (φ) and • to the bulk modulus of the rock forming mineral (Kg). In the SACS-case shear wave information through a DSI-log has proven very valuable for the Gassmann modeling and AVO analysis. Assuming the validity of Gassmann’s formula, and assuming that the three independent mechanical parameters (Ksat , µsat and Kg ) of the rock are known, the impact of fluid saturation can be computed. In order to do this, we need: • to check the validity of Gassmann's formula, • to measure the rock parameters and • to compute the compressibility and density of the saturating mixtures. Note that Gassmann’s formula does not depend on the chemical nature of the gas. Only the thermodynamic properties of the injected gas (-mixture) count. Within SACS a reliable method has been developed for the laboratory verification of Gassmann's formula and parameters by measurement on consolidated samples (Zinszner, 2002). The method is based on the substitution of fluids of various compressibility’s. To preserve the properties of the clay fraction in the sandstone diphasic saturation states have been used. The room dry sample is first saturated with brine. The brine is displaced by viscous oil (non-miscible viscous displacement), and then the viscous oil is displaced by hydrocarbon liquids of varying bulk modulus (e.g. kerosene, hexane, pentane, etc).

24

The P and S wave velocity measurements are performed under pore and confining pressure (up to 70 MPa). This method is very successful when performed on normally consolidated samples, but the experimental difficulties in applying this method to loose sandstone are expected to be large. Similar difficulties are encountered for any petrophysical measurement; permeability, capillary pressure etc, but they are more pronounced for petroacoustics (a careful preservation of the initial rock microstructure is needed). In order to provide the CO2 - methane mixture compressibility (isothermal, isentropic) and density for temperatures and pressures in the range encountered in the reservoir, the SBWR (1995) equation has been applied (a modification by Soave of the 1940 Benedict-Webb-Rubin equation). At the beginning of the SACS study, it was supposed that the methane concentration could be several percent (relative inefficiency of washing process). Actually the methane content appears lower than 1%. The density/compressibility values for a wide range of P, T conditions and for mixtures with CO2 concentration greater than 95% molar, corresponding to the SACS conditions, have been determined. CO2 concentrations less than 95% molar are far from the Sleipner conditions and would require new computations.

Integration of time-lapse seismic with reservoir flow model Time-lapse data may be compared to results from a reservoir simulator with the aim of improving the flow model. Subsequent predictions of reservoir behaviour will then be more accurate. In the SACS case a particular phenomenon occurred due to the thin shale layers acting as temporary vertical CO2 migration barriers that could only be identified on the seismic data with CO2 captured underneath. In other words, assumptions had to be made a priori on the shape, the lateral extent and the continuity of these shale layers for the reservoir simulation model. For that reason in SACS a history match has been performed especially honouring the amount of CO2 at the different depth levels, but only globally (as good as possible) the detailed lateral distribution. Figure 11 shows the synthetic seismics of 2001 created from a realisation of the reservoir simulation model. More information on this topic can be found in Lygren et al., 2002. Other publications are in press.

25

Figure 11: Reservoir simulation model with the corresponding synthetic seismics.

CO2 volume estimation from seismic data The velocity pushdown of reflections beneath the CO2 bubble provides an alternative means of estimating CO2 volume in situ. By interpreting reflectors beneath the bubble on both the 1994 and 1999 surveys it is possible to map the pushdown beneath much of the CO2 bubble. Significant uncertainty arises however because reflections on the 1999 data are locally severely degraded beneath the bubble. A Pre-stack Depth Migration is possibly a way of improving the seismic image in and under the CO2 bubble. However, this is speculative. Supplementary mapping of the pushdown can be effected by performing a cross correlation of the sub-bubble reflections on the 1994 and 1999 surveys, thereby deriving a pushdown time-lag for each grid point. The total amount of pushdown caused by the bubble can be expressed as the individual time-lags at each seismic trace (or bin), summed over the entire anomaly. This is termed the Total Area Integrated Time Delay (TAITD). Such a TAITD could also be derived from a reservoir simulation model providing a verification on the consistency between the seismic interpretation and the reservoir simulation model. Alternatively the amplitudes of the seismic signal can be inverted to a layered model. In the SACS case use can be made of a seismic tuning relation (Arts et al., 2002) in order to estimate the thicknesses of the individual CO2 accumulations. Again such models can be converted to a TAITD for comparison with the other volume estimations. More detailed information will be published in Arts et al. (2002 in press) and Chadwick et al. (2002 in press).

26

Reservoir simulation in SACS: Verifying the seismic and geological interpretations and predicting the long-term fate of CO2

Introduction In geological CO2 sequestration projects pre-injection reservoir simulation should be carried out with a reservoir model which is based on the best available geological data. These simulations can predict the CO2 injection rate that could be maintained, the rise in reservoir pressure caused by the injection, the likely lateral migration of the injected CO2 and the potential for CO2 dissolution into the formation water. Pre-injection reservoir simulation was carried out at Sleipner (Korbol and Kaddour, 1995) but this did not form part of the SACS project, which was established after injection began. The pre-injection reservoir simulation indicated injection of CO2 would be a feasible option from an operational point of view. This was sufficient to allow the project to proceed. Further objectives of reservoir simulation in a CO2 sequestration project are likely to be: 1. Verify and improve the seismic and geological interpretations of the reservoir

around the injection site and re-run simulations of the migration of the injected CO2 during and shortly after the injection period.

2. Use the history matched reservoir model of the area around the injection site to build a large-scale model to predict the long-term fate of CO2.

These objectives require history matching and thus should take place during the monitoring of the CO2 sequestration operation. In the SACS project, two new reservoir models were built to achieve these latter objectives. The first describes the formation near the injection site. It covers an area of approximately 7 km2 and consists of a large number of small grid blocks. This model was iteratively calibrated and adjusted in the light of interpretations of the seismic images of the CO2 accumulations from the repeated seismic surveys performed three and five years after the start of injection. The second model covers an area of 128 km2 and is being used to predict the migration of CO2 over a period of several thousand years under the assumption that there is no migration the through the upper seal. This model has to rely on a coarser grid due to computational constraints.

Calibration of a local reservoir model by use of repeated 3D seismic

Reservoir model The SACS project graphically illustrates how useful repeated 3D seismic surveys can be to calibrate a local reservoir model. Data from pre-injection seismic, well-logs and petrophysical data obtained from laboratory experiments and core analysis were used to build the original local reservoir model of the Utsira Sand (the reservoir formation) near the injection well. However, because the injection well is a near-horizontal well drilled from the Sleipner A platform it did not provide good 3D data on the nature of the whole

27

thickness of the Utsira Sand reservoir at the injection point. Furthermore there are no other wells in the immediate vicinity of the injection site. The majority of the data used to construct the model was obtained from wells which passed through the Utsira Sand beneath, or very close to, the Sleipner A platform, some 3 km from the injection site. At the injection site the Utsira Sand was interpreted to consist of a highly permeable sand body more than 200 m thick intersected by thin horizontal discontinuous shale layers. CO2 is injected close to the bottom of the formation. The shale layers are interpreted to impede its vertical migration and cause the entrapment of the CO2 in large, near-horizontal 'bubbles' within the porous medium of the sand. The barrier layers are either semi-permeable, or have localized spill areas that allow migration of CO2 to the consecutive barrier layers above. The discontinuity and heterogeneity of these shale layers are thought to cause the CO2 to be transported in distinct chimney-like columns that are imaged on the repeat seismic surveys. Only the two upper shale horizons could be mapped from pre-injection seismic i.e. the cap seal of the formation and shale approximately 15 m below the cap (the sand between these two shales is commonly referred to within the SACS project as the 'Sand Wedge'). The other shales were too thin to be mapped from the seismic and were located from the 1999 time-lapse seismic data where the major seismic reflectors were interpreted as CO2 bubbles being retained by the shales. The shale layers were represented in the model by transmissibility modifiers attributed to layers that correspond to those detected by the seismic survey. Reservoir simulation incorporates the predominant driving mechanisms that control the migration of CO2. The model is calibrated by modifying various parameters to achieve history matching and the history-matched model is ultimately adopted to make future predictions. The transmissibility of each shale and the chimney-creating conduits were obtained by adjusting the transmissibility multipliers so that the resulting accumulations under the layers became similar in size to the corresponding seismic reflector. This is an iterative process that is still continuing. Thus the SACS local reservoir model has demonstrated that if a well does not exist at, or very close to, the injection site, as at Sleipner, the initial calibration of the physical conditions and reservoir model may not be ideal. However, if good quality 4D seismic data is available, the reservoir simulation can still be history matched to the seismic interpretation.

Fluid and transport properties Given a hydrostatic pressure gradient, in a thick reservoir such as the Utsira Sand the temperature gradient is the most important parameter that has to be taken into account if fluid properties are to be modelled correctly. Thus we recommend that careful temperature and pressure measurements are made in the reservoir in future CO2-injection projects. The CO2 density in particular will be erroneous if these gradients are not correctly accounted for.

28

In the Utsira Sand, the temperature is thought to vary from about 29°C to 37°C from the top of the formation about 800 m below mean sea level to the injection point at 1040 m depth. The pressure increases downwards through the formation and temperature and pressure have opposite effects on the density, so in practice the density is relatively constant down through the reservoir, at about 700 kg/m3 corresponding to a CO2 viscosity of about 0.06 mPa s. Free CO2 in both liquid or gas phase will give strong reflections on seismic because of the strong contrast in velocity of sound between CO2 and brine. CO2 dissolved in brine will, however, not be visible on seismic because CO2 saturated brine will have approximately the same velocity of sound as under-saturated brine. The solubility of CO2 in brine at the Utsira conditions is approximately 53 kg/m3. Dissolved CO2 could therefore potentially be a significant contribution to CO2 storage in this aquifer, e.g. all of the CO2 injected in this project (1.7·106 Sm3/d) for 25 years would dissolve in a brine “cylindrical” pore volume 1300 m in radius and 200 m tall. In the CO2 plume above the injection point some water will be contacted by CO2 during migration up through the formation. The shales will spread the CO2 over a large area. This will increase the surface of the CO2 phase and increase dissolution. In practice, however, the amount CO2 dissolved during the injection period will be limited because only a small fraction of the brine will be contacted by CO2. Although the geophysical interpretation of the seismic is non-unique, iteration between the geophysical interpretation of the seismic reflections attributed to the injected CO2 and the reservoir simulations showed that good matches between observed and simulated bubble areas could be achieved even if CO2 solubility was completely neglected. From this it can also be concluded that the shale layers do not disperse large amounts of CO2 into small leak streams when it is transported from layer to layer. The CO2 transport must rather be concentrated at localised spill points, curtains, or holes.

Simulation tools Simulations were carried out with two different 3D simulators, Simed II and Eclipse 100. Simed II is a multi-component reservoir simulator which initially was designed for modelling the drainage of methane from coal seams (CBM). The simulator includes a gas phase density calculation using a Peng-Robinson equation of state with a Chien-Monroy correction, and viscosity by the Jossi-Thiel-Thodos method. The simulator was implemented with an option to specify depth-related temperatures for each grid block thus preserving a consistent density versus depth profile. Eclipse 100 is a black-oil simulator that can handle up to four flowing phases. Only the oil and gas phase were used in the SACS simulations. The oil phase was given pVT and phase data corresponding to brine and the gas phase was given properties corresponding to CO2. pVT data, solubility data and viscosities are represented in tables. This allows both solubility properties and density versus depth data to be consistently represented as the pressure variation in the model is dominated by the hydrostatic pressure gradient

29

throughout the simulation. CO2 densities for the pVT table were calculated by an EOS developed by Span and Wagner (1996). The SACS reservoir simulation programme has demonstrated that existing simulators can successfully model CO2 migration provided that the physical properties of the CO2/brine system are well represented. For all time-scales the density difference between brine and CO2 and the viscosities are the dominating fluid parameters.

Results of model calibration Examples of the results of are illustrated graphically as 2D contours of the seismic and simulated bubbles and in 3D graphs (Figures 11-14).

30

Layer 39

6470000

6470500

6471000

6471500

6472000

6472500

437500 438000 438500 439000 439500 440000X

Y

Layer 35

6470000

6470500

6471000

6471500

6472000

6472500

437500 438000 438500 439000 439500 440000X

Y

Layer 32

6470000

6470500

6471000

6471500

6472000

6472500

437500 438000 438500 439000 439500 440000X

Y

Layer 25

6470000

6470500

6471000

6471500

6472000

6472500

437500 438000 438500 439000 439500 440000X

Layer 16

6470000

6470500

6471000

6471500

6472000

6472500

4 3 7 5 0 0 4 3 8 0 0 0 4 3 8 5 0 0 4 3 9 0 0 0 4 3 9 5 0 0 4 4 0 0 0 0

X

Layer 21

6470000

6470500

6471000

6471500

6472000

6472500

437500 438000 438500 439000 439500 440000

X

Y

Layer 19

6470000

6470500

6471000

6471500

6472000

6472500

437500 438000 438500 439000 439500 440000

X

Y

Layer 8

6470000

6470500

6471000

6471500

6472000

6472500

437500 438000 438500 439000 439500 440000

X

Y

Figure 11 Comparison of eight bubble contours observed by seismic after three years of injection (left) and simulated with Simed II (right).

31

Figure 12 An alternative seismic interpretation with only six horizons is mimicked with Eclipse 100. Contours represent seismic while the blue area represent simulation results both corresponding to 5 years of injection

Figure 13 3D simulated smoothed image of the CO2 saturation after 3 years of injection corresponding to the maps with eight horizons in Figure X.1. (Simed II)

32

Top Sand wedge

Top Utsira

Top Sand wedge

Top Utsira

Figure. 14. 3D simulated image (left) after five years of injection compared to seismic images (right) corresponding to the maps with six horizons in Figure 12, (Eclipse 100)

Simulation of the long-term fate of CO2 in a large-scale model One of the main objectives of reservoir simulation in a geological CO2 sequestration project is to make long term predictions of the fate of the injected CO2. The reservoir model constructed for this purpose should include the major features of the local model that control transport of CO2 on the relevant time scale. The fluid model of CO2 and brine must feature correct volumetric data (densities), phase behaviour (solubility) and transport properties (viscosities and diffusion coefficient). In the SACS project, the information from the calibrated local model was extrapolated to build a 3D reservoir model covering an area of 128 km2 to predict the fate of CO2 over a time period of thousands of years. Capillary pressure and relative permeability describing the interaction between the porous media and the fluids were measured in laboratory experiments on Utsira cores. Computational constraints limited the number of grid blocks in the model to less than one million to achieve acceptable computation times. This represents a substantial coarsening of the grid compared to the local model. Preserving the physical consistency of the major transport phenomena in the new grid is a major challenge. In the model the cap rock shales are assumed to provide a capillary seal for the CO2 phase preventing upward migration, but allowing molecular diffusion of CO2 through the overlying strata. The results of the simulations show that most of the CO2 accumulates in one bubble under the cap seal of the formation a few years after the injection is turned off. The CO2 bubble spreads laterally on top of the brine column and the migration is controlled by the topography of the cap seal only.

33

Molecular diffusion is driven by concentration gradients and can usually be neglected in reservoir simulations as it is a slow process compared to other transport processes. It is attenuated due to diminishing concentration gradients, which is a result of the diffusion process itself. In this case, however, diffusion of CO2 from the gas cap into the underlying brine column will have a most pronounced effect. The brine on top of the column, which becomes enriched in CO2, is denser then the brine below due to the special volumetric properties of the CO2-brine system. This creates an instability that sets up convectional currents maintaining a large concentration gradient near the CO2/brine interface, enhancing the dissolution of CO2. This is illustrated in Figure 15. Maps of the bubble as function of time are shown in Figure 16, where the top of the sand wedge is the controlling seal. In these simulations the dissolution of CO2 is neglected. If dissolution is included the bubble will reach a maximum size after probably less than 300 years. After this time dissolution is the dominating effect on bubble extension and the bubble will gradually shrink and finally disappear after less than 4000 years. Thus preliminary results suggest that in the long term (> 50 years) the phase behaviour (solubility and density dependence of composition) will become the controlling fluid parameters at Sleipner. An alternative scenario where Top Utsira Sand (i.e. the top of the sand below the Sand Wedge) is the controlling topography for migration was also simulated. Figure 17 shows that the CO2 will follow a more eastern path. This illustrates how sensitive the migration is to small changes in topography. Top Utsira and the top of the sand wedge are only between 14 and 35 m apart and relatively parallel. The top of the sand wedge dips slightly more towards the south west though, resulting in the large differences between distribution patterns. This test is only presented to illustrate the sensitivity of topography because of it is quite unlikely that Top Utsira will retain any CO2 on long term because of its permeability. Upward molecular diffusion of CO2 through the water-saturated overlying shales can potentially represent an escape path for CO2 into the atmosphere. Along this pathway injected CO2 will not reach the sea floor until several hundred thousand years after the end of injection. This escape mechanism can in practice be neglected.

34

35

8 years 10 years

5 years4 years

1 year

1 year

3 years

Figure 15. Concentration profiles in a 10 x 13.6 m segment just below the CO2 brine contact. From a meta-stable diffusion front (upper left) convectional plumes gradually develop. This convection gives a significant contribution to the dissolution.

8 years

8 years

Figure 16 . Maps of the CO2 bubble migrating under the top of the sand wedge as function of time. CO2 dissolution has been neglected. After 500 years CO2 reaches the boundaries of the model and starts to migrate out of the model.

36

Figure 17. Maps of the CO2 bubble migrating under the Top Utsira as function of time. CO2 dissolution has been neglected. In this case the CO2 follows a much more eastern path than in the case were the top of the sand wedge was controlling the migration.

37

Conclusions and recommendations

Co

ory to date is best simulated with a reservoir

to the large temperature

ust be paid most attention. ic data can be closely related

ver at more than one spill point

gh the shales In all tests the top seal was assumed to be a perfect capillary seal for CO2. This

verified.

mited because the geology of each sequestration site is likely to be highly site specific.

In the f

• For all

rs. In the long term (> 50 years) the phase

behaviour (solubility and density dependence of composition) will become the

•

ation model to the development of the seismic images. This information can be used to

•

ather than faults, the heterogeneities will be a controlling parameter only on short time-scales (< 25 years) even if the CO is injected deep

• The topography and quality of the seal will be the controlling geological

parameter in long-term simulations.

nclusions • Existing simulators (e.g. Simed II and Eclipse 100) are well suited for calibration

of the reservoir model by help of the observed seismic images of the bubbles • Simulation of the migration hist

model with tight horizontal shales with highly permeable holes, spill points or curtains controlling the vertical migration.

• Fluid density relations can be modelled, but duedifference in the aquifer column and the strong temperature dependence of the CO density, the reservoir temperature m2

• The area of the bubbles as interpreted from the seismto the known volume of the accumulated CO2

• CO will spill o2

• Relative mobility effect can play an important role especially to describe the transport throu

•has yet to be

Recommendations It must be emphasised that the lessons that can be learned from the present work areli

ollowing recommendations only the most generic information is presented:

Existing reservoir simulators can be applied to model CO2 migration provided that the physical properties of the CO2/brine system are well represented.time-scales the density difference between brine and CO2 and the viscosities arethe dominating fluid paramete

controlling fluid parameters.

By monitoring the CO2 distributions by repeated 3D seismic, quantitative information describing the formation can be provided by calibrating a simul

build larger reservoir models, which can be used for long-term predictions.

In highly permeable reservoirs where the heterogeneities are dominated by horizontal shales r

2in the formation.

38

Assessing the geochemical effects of CO2 injection It is essential to have a good understanding of the fluid chemistry and mineralogical composition of any potential reservoir and cap rock so as to elucidate their reactivity with CO2. The degree of reactivity between CO2, pore water and minerals will influence the long-term storage potential of the formation. For example, instead of being trapped as a buoyant supercritical CO2 ‘bubble’ (physical trapping), reaction with formation water could trap the CO2 as a dissolved phase (solubility trapping). Furthermore, reaction of this dissolved CO2 with minerals in the host formation could result in pH buffering, enhancing solubility trapping due to the formation of dissolved bicarbonate ions and complexes (ionic trapping). Reaction of dissolved CO2 with certain non-carbonate calcium-rich (or Fe, Mg,…rich) minerals could even trap the CO2 as a solid carbonate precipitate (mineral trapping), essentially immobilising the CO2 for geological time periods. This terminology about trapping mechanisms is derived from Bachu et al. (1994). Depending on the nature and scale of the chemical reactions, the reservoir-CO2 interactions may have significant consequences for the CO2 storage capacity, the injection process, and long-term safety, stability and environmental aspects of CO2 storage (Czernichowski-Lauriol et al., 1996a, b).

The fluid chemistry and mineralogical composition of reservoir and cap rocks is site specific, so it is important to recognise that geochemical investigations need to be carried out on a site-to-site basis. In the following description we aim to draw out general recommendations for assessing the geochemical effects of CO2 storage, with specific examples being based on experience gained during the SACS project. We recommend the following approach:

1. Determination of baseline geochemical conditions prior to CO2 injection 2. Determination of the geochemical impact of injected CO2

2.1 Observations: - from laboratory experiments - from field monitoring - from natural analogues

2.2 Numerical modelling: - to aid interpretation of the above observations - to give predictions over a variety of spatial and temporal scales

Determination of baseline geochemical conditions prior to CO2 injection A good geochemical understanding of the system will require knowledge of the ‘baseline’ conditions of mineralogy and fluid chemistry prior to CO2 injection. Only with this information can changes due to the presence of CO2 be assessed. It is important therefore, that sample acquisition be implemented prior to CO2 injection operations.

39

Baseline geochemistry can be best determined by analysis of suitably preserved borehole core material and pore waters. This should aim to produce data on the chemical makeup of reservoir and cap rock formations prior to any CO2 injection. This should include, as a minimum, for both the reservoir and cap rock formations:

• Mineralogy Mineralogical and chemical characterisation of solid phases Identification of detrital and authigenic phases Specific surface areas Cation exchange capacity (CEC) and exchangeable cations Recommended analytical tools include Optical microscopy, SEM

(scanning electron microscopy), XRD (X-ray diffraction), Electron microprobe analysis, Particle-size analysis, BET (specific surface measurements)

• Transport properties Porosity Absolute permeability and relative permeability Threshold pressure for the CO2-brine-caprock system

• Pore water properties Cations (e.g. Li, Na, K, Mg, Ca, Sr, Ba, Mn, total Fe, Al, Si, total S, and

others as necessary) Anions (e.g. Br-, Cl-, SO4

2-, and others as necessary) pH with corresponding temperature Alkalinity Total inorganic carbon (TIC) Total organic carbon (TOC)

• In-situ conditions of temperature and pressure • Natural fluid flow within the reservoir

Water samples should come from the same location as rock samples in order to relate more closely fluid chemistry to mineralogy.

Reservoir formation Knowledge of the chemical makeup of the reservoir and its properties is required to quantify possible chemical reactions and their rates, together with overall potential storage capacity. Carefully collected formation pore waters from borehole pumping tests are very valuable. For all surface sampling, water flow rate and gas-water ratio as well as non-conservative parameters (e.g. temperature, conductivity, pH, Eh, alkalinity) must be measured on site at the wellhead. This is because surface samples have undergone chemical modifications (mainly degassing, but also possibly cooling and mineral precipitation) and indirect calculations are needed to assess fluid chemistry at depth. Gas composition and water composition are then analyzed at the laboratory.

40

Formation pore waters can also be extracted from core material. However the water sample obtained is not representative of in-situ conditions at depth and additional information on gas-water ratios and gas content has to be found, hopefully from the same well. Moreover pore waters extracted from core material are often contaminated by drilling fluids and corrections have to be done to assess the actual fluid chemistry. A possible solution to this problem could be to obtain pressurized samples that would then be available for detailed analysis in a laboratory. However this requires specific tools and know-how and, although the technology exists and could be used, it is not generally common practice. Core and cuttings material from wells intersecting the reservoir formation are necessary for the mineralogical and petrophysical characterisation of the host reservoir. A special effort should be focused on obtaining both fluid and rock samples from the same location as it is essential to relate fluid chemistry to mineralogy for a good assessment of baseline conditions. The availability of these, especially if from several wells, will further improve characterisation, particularly if vertical and lateral reservoir inhomogeneity is suspected. Determinations from material in the likely CO2 migration pathway (e.g. the top of the reservoir) are of particular importance. If possible, preservation of duplicate samples for long-term storage would provide a resource for future studies. At the start of the SACS study only limited geochemical information and samples were available from the Utsira Sand. This included:

• A single (partial) analysis of Utsira formation water from the Oseberg field approximately 200 km north of Sleipner.

• A 7 m core of Utsira Sand from the Sleipner field (of which 1 m sections of frozen core were supplied to the geochemists).

The core sample allowed for detailed mineralogical analyses and determination of transport properties. However, the core sample was heavily contaminated by drilling fluids, and no useable formation water sample could be obtained from it. Only one borehole terminates in the Utsira at Sleipner (the CO2 injection borehole), and unfortunately no produced pore water samples were available from it.

Although there is a single analysis of Utsira pore water from the Oseberg field, it is limited by the lack of analyses of Al and Si. For predictive modelling, it was therefore necessary to assume that these elements were controlled by saturation with respect to specific minerals – in this case kaolinite and chalcedony. However, during the study, a surface sample of formation water from the Brage field (also about 200 km north of Sleipner) was obtained (but without information on the gas phase) and analysed for a range of elements (including Al and Si). However, the sample was unpreserved

41

(unfiltered and un-acidified) and the Al and Si analyses look problematic. Figure 18 summarizes the geochemical data available at the time of the SACS project. Despite this lack of information and samples, a reasonable assessment of baseline conditions within the Utsira sand was made by combining information from the Sleipner, Oseberg and Brage hydrocarbon fields, and through numerical modelling and ‘blank’ experiments. These laboratory experimental investigations were also designed to provide information on in-situ pore water chemistry, as mentioned later.

Reservoir seal Knowledge of the chemical makeup of the reservoir seal and its transport properties is required to quantify possible chemical reactions and their rates, together with overall sealing efficiency. To determine these properties, a minimum prerequisite is to have core material from the cap rock above the injection point. Core and cuttings material from additional wells will further improve characterisation particularly if vertical and lateral cap rock inhomogeneity is suspected. Determinations from material directly in contact with any likely CO2 ‘bubble’ (i.e. the base of the cap rock) are of particular importance. Knowledge of the sealing capacity of the cap rock is perhaps the key element in assessing and establishing the long-term safety case for CO2 containment. Two aspects are important here; the natural seal (i.e. cap rock), and the man-made seal around breaches in the cap rock (i.e. boreholes). Cap rock core material should be available in sufficient quantities to undertake a detailed suite of analytical tests. Ideally, the core material should be in a location above the likely CO2 migration pathway, or from a demonstrably analogous position. Samples of borehole cement should also be available for testing and analysis.

During the SACS study no cap rock core material was available for study. It was therefore not possible to study its bulk properties and pore water chemistry. However, some drill cuttings were located, and cleaned of drilling fluids. These cuttings were suitable for a limited range of mineralogical analytical techniques (petrography, SEM, XRD). Results from these tests were used to assess sealing capacity through comparison with samples from proven oil/gas field cap rocks. The Krushin grain-size method was also used. The interactions of CO2 with borehole cement were not addressed in this study. A key aspect of any future investigations at Sleipner would be to obtain cap rock core material and samples of borehole cement. The properties of these, and their interactions with CO2, could then be investigated in detail.

Determination of the geochemical impact of injected CO2

The impact of injected CO2 can only really be assessed once there is a sufficiently good understanding of the baseline conditions. Once these have been defined, then changes from them can be more readily identified. There are a variety of approaches that can be used. They combine numerical modelling and observations from laboratory experiments, field monitoring, and natural analogues.

42

Observations from laboratory experiments, field monitoring and natural analogues A range of observations can be achieved at different scales to assess the geochemical impact of injected CO2: from sample to field, from hours to millions of years, from direct study of the selected injection site to indirect study through natural analogues.

The objective is to identify the nature and kinetics of the key physico-chemical reactions between injected CO2, reservoir fluids, reservoir rock and cap rock, and in particular:

• Dissolution reactions • Precipitation reactions • Dehydration reactions • Ion-exchange reactions

Recommended analyses and analytical tools are the same as those listed for the determination of baseline conditions.

Observations from laboratory experiments Direct observations of reactions can be achieved through well-controlled laboratory experiments reacting samples of reservoir rock and cap rock with CO2 under simulated reservoir conditions. Geochemical changes can be followed in detail through direct observation and careful sampling. Such investigations are particularly useful for the study of shorter-term processes. Although limited in scale and timeframe, laboratory experiments have the advantage that they can help to identify the key geochemical reactions on actual rock material under actual reservoir conditions, which is very important as such reactions are known to be highly site-specific. They are also helpful to test the ability of geochemical codes to reproduce the experimental observations before using them to make long term predictions over experimental timescales up to thousands of years.

Both static batch experiments and dynamic core flood experiments are useful. Batch experiments can highlight the potential for reaction of samples of reservoir rock and cap rock when in contact with CO2 at reservoir temperature and pressure, over different timescales. Core flood experiments are aimed at reproducing open systems where the rock is continuously flooded by pore water rich in injected CO2, which represents more closely actual reservoir conditions. They may elucidate inter-relationships between geochemical and hydrodynamic processes, and overall observable reaction may well be higher than in batch experiments.

Ideally, experiments should be undertaken in pairs; one pressurised with CO2 and a ‘blank’ pressurised with an inert gas such as N2. This will help distinguish purely-CO2-related effects from any experimental artefacts, and will thus increase confidence in the assessment of the geochemical impact of injected CO2.

Many different types of experimental equipment can be used to study the reactions between supercritical CO2-porewater-rock reactions. These range from ‘off the shelf’ models to specialized custom-designed pieces of equipment. Some basic requirements of

43

the equipment are that it must be chemically unreactive and able to reproduce in-situ conditions over prolonged timescales. Although dry supercritical CO2 is relatively unreactive towards experimental equipment, it is more reactive when dissolved in water, and especially when this is saline. To minimise corrosion, pressure vessels should be lined with relatively inert materials (e.g. Teflon, gold or silicon carbide), or constructed out of less reactive metals and alloys (e.g. titanium and Hastelloy C-276). Pressure seals and sleeves or membranes must be able to withstand the rigours of supercritical CO2, and Viton O-rings or hydrogenated Nitrile rubber sleeves and O-rings are preferred over cheaper varieties. Likely in-situ temperature and pressure conditions (probably <100 °C and <300 bars) can be reproduced relatively straightforwardly with modern pumps and heaters.

During the SACS project both static and dynamic experiments were assembled. A number of identical static (batch) experiments were undertaken. These simple and relatively low cost experiments used fixed amounts of Utsira Sand and synthetic Utsira pore water, plus fixed temperature and pressure of CO2 (Figure 19). Individual experiments were terminated after different timescales. Detailed analysis of the reaction products provided ‘snapshots’ of reaction progress over a 2 year period. The above experiments were compared to similar experiments pressurized with nitrogen. These latter ‘blank’ experiments were also useful to simulate conditions prior to CO2 injection, and hence helped to fix baseline conditions. Dynamic (flowing) experiments were conducted to investigate how geochemical reactions impacted upon fluid flow and vice-versa. Standard ‘core flood’ equipment was used for several of the tests. However, also used were non-metallic (PEEK) tubes that were joined to create a column of Utsira sand 2.4 m long. Sampling of the experiments presented its own problems, as the aqueous fluids were charged with high pressure CO2. Once these fluids are depressurised, the CO2 comes out of solution (just like when opening a bottle of champagne). Loss of CO2 changes the solution chemistry, possibly causing over saturation with respect to certain minerals, and this may lead to precipitation. Solutions were preserved as soon as possible after sampling so as to minimise any artefacts. Solids had aqueous fluids drained off prior to depressurisation and were washed clean immediately after sampling. This minimised unwanted secondary mineral precipitation. In the dynamic flow experiments, techniques were developed to take fluid samples without interrupting the experiment or disturbing the flow through the sample. The experiments on the Utsira sand have revealed changes in fluid chemistry, associated mainly with dissolution of primary minerals. The experiments pressurised by CO2 led to large and rapid increases in concentrations of Group II metals (and in particular Ca, Sr and Fe), as well as slow and slight increases in silica concentrations. This suggested fast partial dissolution of carbonate phases, while dissolution of silicate or aluminosilicate minerals was a much slower but real process. However, direct evidence from mineralogical observations has never been possible despite the high water-rock ratio used for these experiments (10:1), their relatively long duration (up to 2 years) and the higher temperature (70°C) used for some of them. This is because the reactivity of the Utsira

44

sand was low and changes were below the resolution of the analytical technique or below the natural mineralogical variation within the sand. Observations from field monitoring The most obvious way to obtain direct geochemical information would be by direct sampling of a CO2 injection site. Once baseline conditions are established, longer-term monitoring of the injection process would be required. Access to samples over a range of timescales would be important. This approach would require observation boreholes with repeat fluid sampling to monitor fluid chemical changes. Sidewall coring, or the drilling of boreholes through the CO2 ‘bubble’ could be necessary to obtain samples of rock that had been in contact with CO2 for a variety of timescales. Such an approach would be useful in providing highly relevant ‘real time’ information about a large-scale system. During the SACS project, the lack of observation boreholes and related samples made it impossible to monitor directly the geochemical processes occurring within the Utsira at Sleipner. However such an approach is being used in another industrial CO2 sequestration project - the Weyburn oil field in Canada (http://www.ieagreen.org.uk/weyburn.htm).

Observations from natural analogues This approach utilises relevant information from other sources than the selected site to generate a better understanding of the CO2 injection system. Natural accumulations of CO2 exist in many parts of the world and have many analogous features to any CO2 injection operation, although these may not be exactly comparable. As such, these ‘natural analogues’ can provide much useful information, especially about long-term processes as the CO2 can, in many cases, be proved to have been trapped for thousands or millions of years. Study of natural accumulations of CO2 has the advantage of similar physical size and timescale of reaction. This can build confidence in models that predict likely responses of reservoirs to geological sequestration. However, costly studies including the drilling of boreholes are needed to gain a reasonable understanding of the analogues. Several research projects on natural CO2 accumulations are presently underway in the world, such as the European NASCENT project (http://www.bgs.ac.uk/nascent), the American NASC project (Stevens et al., 2001) and the Australian GEODISC project (http://www.apcrc.com.au/Latest%20Releases/geodisc.htm). One of the objectives of these projects is to study the geochemical effects of CO2 on reservoir rocks and cap rocks, for various geological contexts. Although only analogous to any sequestration system, natural accumulations of CO2 also have the advantage that they are a good way to demonstrate that certain rocks can safely contain CO2 for geological timescales.

Numerical modelling Computer simulations are very useful way to rapidly scope a range of different scenarios. They can predict the effects of CO2 addition to formation pore waters, and the consequent changes in fluid chemistry and reservoir mineralogy. Some codes only deal with static

45

(i.e. not flowing) systems, whereas others couple geochemical reactions and solute transport to produce simulations where mineral precipitation has a direct impact on fluid flow. A wide variety of different geochemical models are available, but they are generally based on the same underlying mathematical expressions. They do differ however, in their emphasis, and the way they handle input and output files. Many come with their own underlying thermodynamic data files, though kinetic data usually have to be supplied by the user. An important benefit of such codes is their ability to make predictions from shorter timescales to the longer timescales, which is essential for the long-term assessment of CO2 storage. However, their output is crucially dependent on the reactions taken into account and the underlying data files (e.g. they will not predict what their data files do not contain) as well as on the reliability of the conceptual model chosen (which requires a good expertise in geochemical processes). Uncertainties resulting from this can be minimised by carrying out a wide range of sensitivity runs and by comparison of modelling results with observations from laboratory experiments, field monitoring and natural analogues, which serve as useful validation test cases.

Within the SACS study, numerical modeling was used to interpret, and hence to better understand the laboratory experiments, based on thermodynamic, kinetic, flow and transport processes. Batch experiments were modeled using geochemical models while core flood experiments were modeled using coupled reactive-transport models. At this stage, 1D simulation was sufficient to describe the core flood experiments. The codes used were EQ3/6 (Wolery, 1995); DIAPHORE (Le Gallo et al., 1998), MARTHE (Thiéry, 1990) and Specific Chemical Simulators (Kervévan and Baranger, 1998; Kervévan et al., 1998) constructed using the ALLAN/NEPTUNIX code generator package (Fabriol and Czernichowski-Lauriol, 1992). For most of the major elements, the predicted trends were in reasonable agreement with the experimental observations on the Utsira sand. However sensitivity calculations were necessary to fit at best the experimental results. This proves that experiments are essential to assess the key site-specific processes relevant to the natural system being studied. As an example, Figure 20 illustrates the behaviour of dissolved calcium. It shows that CO2 injection causes a rapid dissolution of calcium carbonate phases, marked by an increase of dissolved Ca concentrations up to calcite equilibrium.

Conclusions and recommendations A recurring theme is that geochemical reactions are highly site specific, because they depend on the precise mineralogy, fluid chemistry, pressure and temperature of the host formation. They are also strongly time-dependent, due to the wide range of reaction kinetics. As a consequence, a precise characterisation of the baseline conditions is very important. Short term and long term predictions about how the system will evolve once CO2 is injected can be dealt with using numerical models. To reduce uncertainties, a wide range of sensitivity runs are necessary, together with benchmarking the modelling results with observations at different timescales from laboratory experiments, field monitoring and natural analogs.

46

Within the SACS project, the objective of the geochemical investigations was to assess the potential for geochemical reactions between injected CO2, formation water and the Utsira sand, based on direct observations from laboratory experiments under simulated reservoir conditions for timescales up to 24 months. Unfortunately, only limited geochemical baseline data were available within the SACS project. This necessitated the use of certain (logical) assumptions in the design of the experimental programme and in the modelling work. In general, the Utsira sand showed only limited reaction with CO2. Most reaction occurred with carbonate phases (shell fragments), but these were a very minor proportion (about 3%) of the overall solid material. Silicate minerals showed only slow and minor reaction. Then, in terms of geochemical reactions, the Utsira sand would appear to be a good reservoir for storing CO2. However further studies are needed to assess the long term storage behaviour within the Utsira formation. In particular, numerical modelling at reservoir scale should be carried out, such as initiated by Johnson et al. (2001). This implies feedback between reservoir simulations and geochemical modelling. Another key area that still remains highly uncertain is the behaviour of CO2 with the reservoir seal (both cap rock and borehole cement seals). Analysis of borehole core material from the cap rock at Sleipner is the only way to provide sufficiently detailed information on cap rock mineralogy and pore water chemistry. Acquisition of such material should be a priority.

47

Figure 18: Baseline geochemical data from the Utsira Formation available during the SACS project

48

Figure 19: Schematic diagram of the batch reactor used for SACS experiments

Magnetic stirer

CO2

inlet

Utsira sand

PTFE lining

CO2 phase

Aqueous phase

H2O outlet

Filter

49

Figure 20: Behaviour of dissolved Ca concentrations during the SACS batch experiments reacting Utsira sand with CO2 at reservoir temperature and pressure (37°C, 10 MPa). CO2 experiment: pressurised with CO2 Blank experiment: pressurised with N2

Total dissolved calcium

0.0E+00

5.0E-03

1.0E-02

1.5E-02

2.0E-02

2.5E-02

3.0E-02

3.5E-02

4.0E-02

4.5E-02

5.0E-02

0 100 200 300 400 500 600 700 800

Time (days)

Con

cent

ratio

n (m

ol/k

g H

2O)

Measured Ca (blank)Measured Ca (CO2)Modelled Ca (blank)Modelled Ca (CO2)

Calcium variation per mineral versus time.

-4.0E-02

-3.5E-02

-3.0E-02

-2.5E-02

-2.0E-02

-1.5E-02

-1.0E-02

-5.0E-03

0.0E+00

5.0E-03

1.0E-02

0 100 200 300 400 500 600 700 800

Time (days)

Del

ta C

a (m

ol/k

gH2O

)

Ca calcite (blank)Ca dis-dolomite (blank)Ca Ca-montmor (blank)Ca calcite (CO2)Ca dis-dolomite (CO2)Ca Ca-montmor (CO2)

50

References Arts, R., Brevik, I., Eiken, O., Sollie, R., Causse, E., & van der Meer, B. 2000: Geophysical methods for monitoring marine aquifer CO2 storage - Sleipner experiences. In: D.J. Williams et al. (eds.): Greenhouse Gas Control Technologies, CSIRO Publishing, Collingwood, Australia pp. 366 – 371.

R. Arts, O. Eiken, A. Chadwick, P. Zweigel, L. van der Meer, B. Zinszner, 2002. Monitoring of CO2 injected at Sleipner using time lapse seismic data. GHGT-6 Conference, Kyoto 2002.

Arts, R, Eiken, O., Chadwick, R.A., Zweigel, P., van der Meer, L.G.H., Kirby, G. (submitted 2002): Seismic monitoring at the Sleipner underground CO2 storage site (North Sea). Submitted to Geol. Soc. Spec. Publ. on underground CO2-storage. Bachu S., Gunter W.D. and Perkins E.H. (1994). Aquifer disposal of CO2: hydrodynamic and mineral trapping. Energy Conversion and Management, 35, 269-279.

Baklid A, Korbul R & Owren G. 1996. Sleipner Vest CO2 disposal, CO2 injection into a shallow underground aquifer. SPE paper 36600, presented at 1996 SPE Annual Technical Conference and Exhibition, Denver Colorado, USA, 6-9 October 1996. Brevik, I., Eiken, O., Arts, R.J., Lindeberg, E., & Causse E. 2000: Expectations and results from seismic monitoring of CO2 injection into a marine aquifer. 62nd EAGE meeting, Glasgow, paper B-21.

Chadwick, R.A., Arts, R, Eiken, O., Kirby, G.A., Lindeberg, E., & Zweigel, P. (submitted 2002): 4D seismic imaging of a CO2 bubble at the Sleipner Field, central North Sea. Submitted to Geol. Soc. Spec. Mem. on 4D seismic. Czernichowski-Lauriol I., Sanjuan B., Rochelle C., Bateman K., Pearce J., Blackwell P. (1996a). 'Analysis of the geochemical aspects of the underground disposal of CO2: scientific and engineering aspects'. In: Deep Injection disposal of hazardous and industrial waste, ed. John A. Apps and Chin-Fu Tsang, 565-583, Academic Press.

Czernichowski-Lauriol I, Sanjuan B, Rochelle C, Bateman K, Pearce JM & Blackwell P. (1996b). Inorganic geochemistry. In Final report of the Joule II project No. CT92-0031: The Underground Disposal of Carbon Dioxide, ed. S Holloway, 183-276, British Geological Survey, Keyworth, Nottingham, UK.

Czernichowski-Lauriol I et al 2002. SACS2 Final Technical Report in press

51

Eiken, O., Brevik, I., Arts. R., Lindeberg, E., & Fagervik, K. 2000: Seismic monitoring of CO2 injected into a marine aquifer. SEG Calgary 2000 International conference and 70th Annual meeting, Calgary, paper RC-8.2.

Fabriol, H., 2001. Feasibility study of microseismic monitoring (Task 5.8). BRGM Commissioned Report BRGM/RP-51293-FR (Confidential). Fabriol, R. & Czernichowski-Lauriol, I. (1992) - A new approach to geochemical modelling with an integrated simulator generation system. In: Water-Rock Interaction, Y.K. Kharaka and A.S. Maest (Eds.). Balkema Publ., pp. 213-216. Gassmann, F., 1951. Uber die Elastizitat poroser Medien. Vier, der Natur, Gesellschaft in Zurich, 96, 1-23. Kervévan, C. and Baranger, P. (1998) - SCS: Specific Chemical Simulators dedicated to chemistry-transport modelling. Part I – Design and construction of an SCS. Goldschmidt Conference, Toulouse, 29th August-3rd September. In: Min. Magazine, 62A, pp. 771-772. Kervévan, C., Thiéry, D. and Baranger, P. (1998) - SCS: Specific Chemical Simulators dedicated to chemistry-transport modelling. Part III – Coupling of SCS with the hydro-transport modelling software MARTHE. Goldschmidt Conference, Toulouse, 29th August-3rd September. In : Min. Magazine, 62A, pp. 773-774. Korbul R & Kaddour A. 1995. Sleipner Vest CO2 disposal – injection of removed CO2 into the Utsira Formation. Energy Conversion and Management, Vol. 36, No. 6-9, 509-12. Krushin, J T. 1997. Seal capacity of non smectite shale. In: Surdam, R C. (ed.) Seals, Traps, and the Petroleum System. American Association of Petroleum Geologists, Memoir 67, 31-47.

Le Gallo, Y., Bildstein, O., Brosse, E. (1998) - Modelling Diagenetic Changes in Permeability, Porosity and Mineral Compositions with Water Flow. In: J. Hydrology Spec. Publ. on "Reaction-Transport Modelling", C. Steefel (Ed.). Elsevier Sciences. Vol 209, Issue 1-4, pp366-388. Liu, E., Li, X.Y. & Chadwick, R.A. 2001. Saline Aquifer Storage: A Demonstration Project at the Sleipner Field. Work Area 5 (Geophysics) – Multi-component seismic monitoring CO2 gas cloud in the Utsira Sand: a feasibility study. BGS Commissioned Report CR/01/064 (Confidential). M. Lygren, E. Lindeberg, P. Bergmo, G.V. Dahl, K.A. Halvorsen, T. Randen, and L. Sonneland (2002). History matching of CO2 flow models using seismic modelling and

52

53

time-lapse data. In Expanded Abstr., Int. Mtg., Soc. Exploration Geophys., Salt Lake City, UT, USA, Oct. 2002. Soave G., A non cubic equation of state for the treatment of hydrocarbon fluids at reservoir conditions, Ind. Eng. Chem. Res., 1995, 34, p. 3981-3994.

Span, R., Wagner, W., 1996 “A New Equation of State for Carbon Dioxide Covering the Fluid Region from the Triple-Point Temperature to 100 K at Pressures up to 800 MPa”, J. Phys. Chem. Ref. Data, Vol. 25, No. 6, pp. 1509 - 1596,. Stevens S.H., Pearce J.M., Rigg A.J. (2001). ‘Natural analogs for geologic storage of CO2: an integrated global research program’. First National Conference on Carbon Sequestration, U.S. Department of Energy, May 15-17, 2001, Washington, D.C. (http://www.netl.doe.gov/publications/proceedings/01/carbon_seq/6a1.pdf) Thiéry D. (1990) - Logiciel MARTHE : Modélisation d’Aquifères par un maillage Rectangulaire en régime Transitoire pour le calcul Hydrodynamique des Ecoulements. Rapport BRGM 4S/EAU n° R32210, 200 pp. Williamson, J.P., Chadwick, R.A., Rowley, W.J., Eiken, O., 2001. Gravity monitoring of the CO2 bubble. BGS Commissioned Report CR/01/063 (Confidential). Wolery T.J. (1995). EQ3/6, a software package for geochemical modelling of aqueous systems: package overview and installation guide (version 7.2b). Lawrence Livermore National Laboratory, Livermore, California. Zinszner, B., 2002. Petroacoustics of poorly consolidated reservoir rocks saturated with CO2/ Methane/Brine Mixtures. IFP Commissioned Report (Confidential).